Академический Документы
Профессиональный Документы
Культура Документы
pdf
A review on the platform design, dynamic modeling and control of hybrid UAVs.pdf
Electric vehicle fleet management in smart grids A review of services.pdf
Integrated Vehicle Dynamics Control.pdf
Review of course keeping control system for unmanned surface vehicle.pdf
Review of Dynamic Research for NVH Problems Related to Automotive Driveline.pdf
Review of research on agricultural vehicle autonomous guidance.pdf
Review of the State of Development of Advanced Vehicle Control Systems AVCS (1).pdf
Special issue on the 4th International Tyre Colloquium.pdf
State of the art survey active and semi-active.pdf
The vehicle routing problem A taxonomic review.pdf
Tire Model Application and Parameter Identification-A Literature Review.pdf
Validation of vehicle dynamics simulation models a review.pdf
Various Approaches for Driver and Driving Behavior Monitoring.pdf
A review of active yaw control system for vehicle handling and stability enhancement.pdf
A review of design and control of automated guided.pdf
A review of experimental techniques for NVH analysis on a.pdf
A review of fuzzy methods in automotive.pdf
A review of human sensory dynamics for application to models.pdf
A review of system dynamics models applied in transportation.pdf
Jurnal
Teknologi
Full paper
Full paper
Jurnal
A Review
Teknologi
Instrumentation & Automation Dept., Faculty of Electrical Engineering, UniversitiTeknikal Malaysia Melaka, Melaka
of Control and Mechatronics, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310 UTM Johor Bahru, Johor Malaysia
cIndustrial Automation Section, Universiti Kuala Lumpur Malaysia France Institute
d
Temasek Laboratory, National University of Singapore
bDepartment
*Corresponding author:khairiaripin@utem.edu.my,khairi32@live.utm.my
Article history
Abstract
A review study on integrated active steering and braking control for vehicle yaw stability system
is conducted and its finding is discussed in this paper. For road-vehicle dynamic, lateral dynamic
control is important in order to determine the vehicle stability. The aw stability control system is
the prominent approach for vehicle lateral dynamics where the actual yaw rate and sideslip should
be tracked by the controller close to the desired response. To improve the performance of yaw
stability control during steady state and critical driving conditions, a current approach using active
control of integrated steering and braking could be implemented. This review study discusses the
vehicle models, control objectives, control problems and propose control strategies for vehicle
yaw stability control system. In the view of control system engineering, the transient performances
of tracking control are essential. Based on the review, this paper discusses a basic concept of
control strategy based on the composite nonlinear feedback (CNF) and sliding mode control
(SMC) whichcan be proposed for integrated active steering and braking control in order to
improve the transient performances of the yaw rate and sideslip tracking control in the presence of
uncertainties and disturbances.
Graphical abstract
Fx1
d
Fx2
Fy2
Fy1
vx
lf
r Mz
vy
Fx3
Fy3
Fx4
lr
Fy4
Keywords: Vehicle yaw stability; active steering control, active braking control, lateral dynamics
control
1.0 INTRODUCTION
This paper presented a review study on integrated active steering
and braking control for yaw stability control system of roadvehicles. The main objective of this paper is to provide an
overview and comprehensive understanding of the main elements
in yaw stability control system and its control strategies in
improving vehicle handling and stability performances.
In vehicle dynamics studies, vehicle stability is determined
by the lateral dynamic motion where yaw stability control system
is one of the prominent approaches for lateral dynamics control
and it can be controlled by using steering and braking subsystem.
In yaw stability control strategy, it is required that designed
controller could control the actual vehicle yaw rate and sideslip
angle with fast responses and good tracking capability in
following the desired responses. Improper commands to steering
and braking subsystems by the driver during critical driving
manoeuvres can cause the vehicle to become unstable and could
lead to accidents. Therefore, an active control is essential to assist
106
M.K Aripin et al. / Jurnal Teknologi (Sciences & Engineering) 71:2 (2014) 105-111
Fx1
( F y1 F y 2 ) cos f F y 3 F y 4
Yaw motion:
I z r l f ( F y1 cos f F y 2 cos f Fx1 sin f
Fx 2 sin f ) l r ( F y 3 F y 4 ) M z
(2)
Fx2
Fy2
vx
lf
r Mz
vy
Fx3
Fy3
Fx4
lr
Fy4
d
Figure 1 Full vehicle model
description
vehicle mass
distance from front axle to centre of gravity
distance from rear axle to centre of gravity
width track
moment of inertia
front tire cornering stiffness
rear tire cornering stiffness
vehicle speed
yaw moment
front wheel steer angle
r
vx
vy
Fxi
Fyi
vehicle sideslip
yaw rate
longitudinal velocity
lateral velocity
longitudinal tire forces
lateral tire forces
where
Mz
Fy1
(1)
(5)
(4)
107
M.K Aripin et al. / Jurnal Teknologi (Sciences & Engineering) 71:2 (2014) 105-111
(6)
a12 a 21 a11 a 22
a12 (a 21e1 a11e2 )
(10)
tr
a12 a 21 a11 a 22
.k r
a 22 (a 21e1 a11e2 )
(11)
k 0
kr
(12)
v
l (1 k us v 2 )
(13)
also that a11 , a12 , a 21 and a 22 in equation (10) and (11) are as
described in equation (6).
shown in Table 1. The input of the model is the front wheel steer
angle f while the outputs or controlled variables are vehicle
Fyr
v
r
Fyf
Mz
lr
lf
rd
v
(l f l r ) k us v 2
. f
(7)
d t
rd 0
where
0
d t
1 r
k
d r
tr
tr
(9)
108
M.K Aripin et al. / Jurnal Teknologi (Sciences & Engineering) 71:2 (2014) 105-111
fd
desired
vehicle model
d , rd
e1 d
e2 r rd
c
controller
M z
,r
actual vehicle
model
Iz
I zv
Cf
Cf
0
c mv
mv
fd
C l
f f 1 M z C f l f
Iz
Iz
I z
(14)
109
M.K Aripin et al. / Jurnal Teknologi (Sciences & Engineering) 71:2 (2014) 105-111
s( z, t ) : c T e(t ) c1
To realize this concept of variable damping ratio, the CNF
control law is designed toinclude the linear and nonlinear
feedback laws as described in the following equations;
u uL uN
(15)
u Fx Gr (r , y) B P( x x e )
where F is feedback matrix, G is a scalar matrix, B is input
matrix.If P>0 is a solution of the Lyapunov equation and (r , y)
is a nonlinear function which is chosenas follows;
(r , y) e ( y r )
( r , y ) e
(r , y )
(16)
y r
1 e 1
(17)
(e
(1( y y0 ) /(r y0 )) 2
e )
(18)
e (t )
I m 1
e 2 (t )
(19)
where
T
c1 : F (r , y)a12
P
(20)
110
M.K Aripin et al. / Jurnal Teknologi (Sciences & Engineering) 71:2 (2014) 105-111
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
111
M.K Aripin et al. / Jurnal Teknologi (Sciences & Engineering) 71:2 (2014) 105-111
Fig. 1.
I. I NTRODUCTION
During the last two to three decades, Unmanned Aerial
Vehicles (UAVs) have experienced a tremendous development. For now, miniature UAV platforms are dominated by
two main types, i.e., fixed-wing conventional aircraft and
Vertical Take-Off and Landing (VTOL) aircraft, and each
type has its own inherent limitations on such as flexibility,
payload, endurance, and etc. A new and promising trend is
to develop fixed-wing VTOL UAV or the so-called Hybrid
UAV, which can inherit the advantages of both and thus have
the ability of vertical take-off and landing as well as high
cruising speed and enhanced endurance. This enables the
possibility of performing wider range of missions or same
missions with better performance.
Indeed, integrating the advantages of fixed-wing and
VTOL aircraft has long been a concern for many aerospace
and aviation industries. Over the years, there have been
several attempts to build manned hybrid aircraft such as
Bell Boeing V-22 Osprey, Vertol VZ-2, Sikorsky X-wing,
Convair XFY Pogo and Harrier GR7 as shown in Fig. 1(a)
to 1(e), respectively [1][2][3][4][5]. Some of the attempts did
succeed and the aircraft are still operating up to the moment
such as V22-Osprey and Harrier GR7. Nevertheless, within
the last four years, the concept invaded the UAV field as a
number of research groups documented their pioneer work
in literature and a couple of companies even commercialized
the idea. It is believed that the hybrid UAVs will be having
a bright future and will promptly dominate the miniature
UAV market. Still in its infancy, there is a huge space for
978-1-4799-6009-5/15/$31.00 2015 IEEE
Fig. 2.
Fig. 3.
Fig. 5. Block diagram of the flight dynamics model of the hybrid aerial
vehicles.
Fig. 4.
TABLE I
R EPRESENTATIVE MODELING
Aircraft Type
RESEARCH WORK .
Tilt-Rotor
[10], [41], [42], [43], [44], [45], [46], [47], [48], [49]
Tilt-Wing
CSTT
[32], [52], [53], [54], [55], [56], [57], [58], [59], [60],
[61], [62], [63], [64], [65]
DTTT
[35]
1
2
tf
(1)
x(t) = Ax(t) + Bu
(2)
t0
subject to
(3)
x
zi
= f0 (x) + g0 (x)z1
= 8 fi (z1 , z2 , . . . , zn ) + gi (z1 , z2 , . . . , zn )zi+1
(4)
(5)
TABLE II
C ONTROL L AWS C LASSIFICATION OF THE HYBRID UAV S .
Control Law
Advantages
Disadvantages
Applications
PID
LQR
Handles complex dynamic systems and multiple actuators, robust w.r.t process uncertainty,
asymptotically stable for controllable systems,
very large stability margins to errors in the loop
Backstepping
Very robust for external disturbances and irregular parameter uncertainties, deals with all
the states of the system and accounts for the
nonlinearities
GainScheduling
Allows easy understanding and simple implementation of the control laws over the full
flight envelope
Time inefficient
NDI
x1
x2
0
.. .. ..
. . .
(6)
=
+
u
xn1 xn 0
xn
b(x)
a(x)
A technical overview of the hybrid UAVs has been provided in this paper. The common platform design types and
technical details are introduced first. The modeling is then
explained in terms of model structure, and representative
work on the hybrid UAV dynamics modeling is comprehensively addressed. Representative flight control strategies
implemented in the hybrid UAVs are then discussed and compared in terms of theory, linearity and implementation. The
review presented in this paper is expected to be informative
(7)
R EFERENCES
[1] B. Defense, V-22 osprey, Space and Security.
[2] D. C. Dugan, Thrust control of vtol aircraft part deux, in the 5th
Decennial AHS Aeromechanics Specialists Conf, January 2014.
[3] B. Handy, Harrier gr7, Royal Air Force Aircraft and Weapons, pp.
89.
[4] U. S. N. C. Newsletter, Rollout week, 2009.
[5] J. Richmond, Its a helicopter! its a plane, Military Aerospace
Technolgy, High Technology, pp. 68 69, 1985.
[6] M. Streetly, IHS Janes all the world aircraft: Unmanned 2013-2014.
IHS, 2013.
[7] Mini panther fixed wing vtol mini uas,. IAI Panther. [Online].
Available: http://www.iai.co.il/2013/35673-41637-en/IAI.aspx
[8] U. Ozdemir, Y. Aktas, A. Vuruskan, Y. Dereli, A. Tarhan,
K. Demirbag, A. Erdem, G. Kalaycioglu, I. Ozkol, and G. Inalhan,
Design of a commercial hybrid vtol uav system, Journal of
Intelligent & Robotic Systems, vol. 74, no. 1-2, pp. 371393, 2014.
[Online]. Available: http://dx.doi.org/10.1007/s10846-013-9900-0
[9] Y. O. Aktas, U. Ozdemir, Y. Dereli, A. F. Tarhan, A. Cetin, A. Vuruskan, B. Yuksek, H. Cengiz, S. Basdemir, M. Ucar et al., A low
cost prototyping approach for design analysis and flight testing of
the turac vtol uav, in 2014 International Conference on Unmanned
Aircraft Systems (ICUAS). IEEE, 2014, pp. 10291039.
[10] A. Vuruskan, B. Yuksek, U. Ozdemir, A. Yukselen, and G. Inalhan,
Dynamic modeling of a fixed-wing vtol uav, in 2014 International
Conference on Unmanned Aircraft Systems (ICUAS). IEEE, 2014,
pp. 483491.
[11] S. Carlson, A hybrid tricopter/flying-wing vtol uav.
American
Institute of Aeronautics and Astronautics.
[12] Firefly6. Birds Eye View. [Online]. Available: http://www.
birdseyeview.aero/products/firefly6
[13] M. Hirschberg, Project zero: The exclusive story of agustawestlands
all electric technology incubator, Vertiflite, vol. 59, no. 3, p. 10=14,
May-June 2013.
[14] O. Tekinalp, T. Unlu, and I. Yavrucuk, Simulation and flight control
of a tilt duct uav, in 2009 AIAA Modeling and Simulation Technologies Conference, Chicago, IL, 2009, pp. 1013.
[15] J. Holsten, T. Ostermann, and D. Moormann, Design and wind tunnel
tests of a tiltwing uav, CEAS Aeronautical Journal, vol. 2, no. 1-4,
pp. 6979, 2011.
[16] T. Ostermann, J. Holsten, Y. Dobrev, and D. Moormann,
Control Concept of a Tiltwing UAV During Low Speed Manoeuvring.
Edinburgh, UK: Optimage Ltd., 2012, 1 CD-ROM. [Online].
Available: http://publications.rwth-aachen.de/record/97342
[17] J. Holsten, T. Ostermann, Y. Dobrev, and D. Moormann, Model
Validation of a Tiltwing UAV in Transition Phase Applying Windtunnel
Investigations. Edinburgh, UK: Optimage Ltd., 2012, 1 CD-ROM.
[Online]. Available: http://publications.rwth-aachen.de/record/116436
[18] E. Cetinsoy, E. Sirimoglu, K. T. Oner, C. Hancer, M. Unel, M. F. Aksit,
I. Kandemir, and K. Gulez, Design and development of a tilt-wing
uav, Turkish Journal of Electrical Engineering & Computer Sciences,
vol. 19, no. 5, pp. 733741, 2011.
[19] K. T. Oner, E. Cetinsoy, M. Unel, M. F. Aksit, I. Kandemir, and
K. Gulez, Dynamic model and control of a new quadrotor unmanned
aerial vehicle with tilt-wing mechanism, 2008.
[20] D. K. Koji Muraoka, Noriaki Okada, Quad tilt wing vtol uav:
Aerodynamic characteristics and prototype flight, in AIAA Infotech@Aerospace Conference. American Institute of Aeronautics
and Astronautics, 2009.
[21] J. Dickeson, D. Miles, O. Cifdaloz, V. Wells, and A. Rodriguez,
Robust lpv h gain-scheduled hover-to-cruise conversion for a tiltwing rotorcraft in the presence of cg variations, in 2007 46th IEEE
Conference on Decision and Control, Dec, pp. 27732778.
[22] High speed tilt-rotoer ousts x-wing project, Flight Internation Technology, January 1988.
[23] J. T. McKenna, One step beyond, rotor wing, February 2007.
[24] A. V. Clara and S. Redkar, Dynamics of a vertical takeoff and
landing (vtol) unmanned aerial vehicle (uav), International Journal
of Engineering Research & Innovation, vol. 3, no. 1, 2011.
814
815
Department of Electrical Engineering, Technical University of Denmark, Frederiksborgvej 399, Building 776, room 03, 4000 Roskilde, Denmark
EDF Lab Clamart, Dpartement MIRE, Avenue du Gnral de Gaulle, 92141 Clamart Cedex, France
c
GECAD - Knowledge Engineering and Decision Support Research Center - Polytechnic of Porto (IPP), R. Dr. Antnio Bernardino de Almeida, 431, 4200-072
Porto, Portugal
b
art ic l e i nf o
a b s t r a c t
Article history:
Received 28 April 2015
Received in revised form
30 July 2015
Accepted 11 December 2015
Available online 29 December 2015
Electric vehicles can become integral parts of a smart grid, since they are capable of providing valuable
services to power systems other than just consuming power. On the transmission system level, electric
vehicles are regarded as an important means of balancing the intermittent renewable energy resources
such as wind power. This is because electric vehicles can be used to absorb the energy during the period
of high electricity penetration and feed the electricity back into the grid when the demand is high or
in situations of insufcient electricity generation. However, on the distribution system level, the extra
loads created by the increasing number of electric vehicles may have adverse impacts on grid. These
factors bring new challenges to the power system operators. To coordinate the interests and solve the
conicts, electric vehicle eet operators are proposed both by academics and industries. This paper
presents a review and classication of methods for smart charging (including power to vehicle and
vehicle-to-grid) of electric vehicles for eet operators. The study rstly presents service relationships
between eet operators and other four actors in smart grids; then, modeling of battery dynamics and
driving patterns of electric vehicles, charging and communications standards are introduced; after that,
three control strategies and their commonly used algorithms are described; nally, conclusion and
recommendations are made.
& 2015 Elsevier Ltd. All rights reserved.
Keywords:
Electric vehicles
Fleet operator
Optimization and control strategies
Smart charging
Vehicle-to-grid
Contents
1.
2.
3.
4.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Service dependent aggregation and its facilitator eet operator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.
Role of electric vehicle eet operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.
Service relationships between eet operators and other actors in a smart grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1.
Providing ancillary services to transmission system operator (TSO) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2.
Providing ancillary services to distribution system operator (DSO) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3.
Providing storage services to renewable energy source supplier (RES). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.4.
Providing charging cost minimization service to electric vehicle owner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.5.
Analysis of the control objectives of eet operator's smart charging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Smart charging infrastructure and modeling of EV battery dynamics and driving pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.
Battery modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.
Charging standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.
Communications standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.
Driving pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Control strategies of eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.
Centralized control of electric vehicle eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
http://dx.doi.org/10.1016/j.rser.2015.12.014
1364-0321/& 2015 Elsevier Ltd. All rights reserved.
1208
1209
1209
1209
1209
1210
1211
1212
1212
1213
1213
1213
1213
1214
1214
1214
1208
4.2.
Transactive control of electric vehicle eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.
Price control of electric vehicle eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.
Discussion on integrating the control strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5. Mathematical modeling and control algorithms: centralized control of eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.
Linear programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.
Quadratic programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.
Dynamic programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.
Mixed-integer linear programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.
Mixed-integer non-linear programming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.6.
Stochastic programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7.
Robust optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.8.
Heuristic and meta-heuristic algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.9.
Model predictive control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.10. A summary of the presented algorithms with three types of criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6. Mathematical modeling and control algorithms: transactive control of eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7. Mathematical modeling and control algorithms: price control of eet operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8. Conclusion and recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.1.
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.
Recommendations on future research directions in the area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Introduction
To achieve the European energy roadmap 2050 [1] The EU is
committed to reducing greenhouse gas (GHG) emissions to 80
95% below 1990 levels by 2050 in the context of necessary
reductions by developed countries as a group. Carbon emission
reduction target are also set up in other countries such as China,
United States, and Korea. The corresponding inuencing factors
studies [26] show that power industry and transportations are
the major contributing factors to CO2 emission. For example, the
power industry was responsible for around 30% of GHG in EU-27
in 2011. The second sector with more GHG emissions in 2011 were
the transports with a share of 20.3% [7]. Thus, the decarbonization
of two main activities including power industry and transportations is required.
To reduce the GHG emissions, in power systems, renewable
sources such as wind and solar power are widely adopted [8,9]. In
the transport sector, electric vehicles (EVs) are important means to
assure the GHG emission reduction goals [4,10]. However, the
increasing investments in renewable power bring operational
challenges into the power systems due to the intermittent
resources. To cope with the new challenges, in smart grid [11,12],
EVs are commonly recognized as one solution in addition to their
environmental benets in the transport sector. They can be utilized to balance power uctuations caused by the high penetration
of intermittent renewable energy sources [13]. However, a largescale integration of EVs also means new loads to electric utilities,
and undesirable congestions and voltage problems may exist in
the distribution network during the charging process [14]. All
these factors bring new challenges to power system operators. As
an outcome, smart charging solutions (including power to vehicle
and vehicle-to-grid (V2G)) are needed in order to make EVs an
asset to the smart grid rather than a merely traditional load.
Much research has been done to capture the benets of electric
vehicles as well as to address the conicting challenges. It is
concluded from the literature that a new business entity, namely
an EV eet operator (FO) has been widely proposed to exploit the
new business opportunities by providing the multiple services of
EVs to system operators. Alternative names for an EV FO are used
such as EV virtual power plant, EV aggregator, EV charging service
provider or EV service provider. The new entities [15,16] could be
independent or integrated in an existing business function of the
energy supplier or distribution system operator (DSO).
1214
1215
1215
1216
1216
1216
1217
1217
1217
1218
1218
1219
1219
1219
1221
1222
1223
1223
1224
1224
modeling of EV battery and driving pattern, charging and communication standards. Three control strategies including centralized control, transactive control and price control are described
in Section 4. Commonly used algorithms in the centralized control,
transactive control and price control of smart charging of EVs are
presented in Sections 57, respectively. Section 8 concludes the
paper with some suggestions for future research.
1209
1210
Fig. 1. The services relationships between eet operator and other actors in a smart grid.
Table 1
Characterization of the literature by the frequency regulation market.
Service name
Short description
Regulation up and down A reserve that stabilizes the frequency in the matter of seconds of
minutes
Spinning reserve
Release the primary reserve and restore the frequency to 50 Hz
Non-spinning reserve
A manual reserve that releases the other two reserves
Overall, the above studies have pointed out that EVs can have a
high contribution to the regulation and spinning reserve services,
especially in the regulation up and down services. These services
require a quick response from the resources (e.g. generators,
demand response programs and EVs) that can be called for a few
minutes. In addition, an adequate telecommunication infrastructure is required to support the ancillary services responding
from the resources. Since EV batteries have characteristic of rapid
responses, EVs are most appropriate for the regulation up and
down. Additionally, the highest price offered in regulation up and
down services can bring more prots for the EV owners. Regarding
spinning reserve, it is also mentioned as a good service for EVs
participation, but with lower revenue, higher duration and higher
availability than regulation service. The spinning service requires
that resources can have a synchronized reserve available to be
called in a typical duration of minutes (e.g. 10 min) until 1 h. With
respect to the non-spinning reserve it is pointed out by the literature that EVs will not be suitable for this service due to its
higher duration (e.g. hours) and lower prices. However, accelerated battery degradation might be the main challenge for a complete implementation of EVs participation in the ancillary services.
Table 1 gives an overview of the literature by the different kind of
frequency regulation services (regulation, spinning and nonspinning reserves).
Yes
[2833]
Yes
No
[28,29,33]
N.A.
Uncoordinated charging;
Smart charging;
Smart charging and discharging (i.e. bidirectional V2G).
An uncoordinated charging occurs when EVs connect to the
distribution network and start immediately to charge until the
batteries reach their capacity. The smart charging consists in an
actor (e.g. DSO or EVs eet operator) controlling the time and
power of the EVs charging. The V2G control is similar to the smart
charging and it has the capability of providing power to the grid.
Studies concerning the charging rate control of EVs and their
effect on the distribution network are dated back to the early
1980s [34]. Heydt [34] argued that load management should be
deployed to ease peak loading by the DSO, which is measured in
term of load factor improvement. In 1993, Rahman and Shrestha
[35] indicated that even low penetration levels of EVs can create
1211
Table 2
Characterization of the literature by the ancillary services in the DSO.
Service name
Short description
Congestion prevention Intelligent control of the charge and discharge of EVs in order to avoid grid congestion,
minimize load peaks and reduce power losses.
Voltage regulation
Intelligent control of the charge and discharge of EVs to improve the voltage prole
avoiding the increase of voltage drops.
References
Yes
[3437]
Yes
[26,3840]
1212
Table 3
Characterization of the literature by the RES service.
Service name
Short description
Renewable integration Evaluate the share of renewables that is possible to have in the power system with Yes
a large scale of EVs.
Storage device
Large mobile storage capacity to mitigate the intermittent behavior of renewables. Yes
References
[44,45,50,51]
[4649]
Table 4
Characterization of the literature by the EV owner service.
Service name
Short description
Expected asset in
the market (e.g.
Danish and
European)
References
Charging
management
Charging and discharging
management
Charging management
aiming at lower cost
Charging and discharging management aiming at maximum prot
Yes
[31,52,53]
Yes
[33,54]
1213
4. Mode 4: fast charging using an external charger in DC. Two submodes of operation are considered for this mode, namely, the
DC level 1 (voltage lower than 500 V, current lower than 80 A,
power at 40 kW) and the DC Level 2 (voltage lower than 500 V,
current lower than 200 A, power at 100 kW) [56,59].
It is important to mention that in charging Mode 1, there is no
communication between the EVs and the charging point through
the connection system. In Mode 2 and 3, a control pilot communication can be included allowing the control of the charging rate.
In charging Mode 4, it is mainly used for the fast charging. A
communication system is included in charging Mode 4 that allows
battery charging management. In addition, in Mode 2, 3, and 4,
wireless communication systems can be used to communicate
with electric vehicles and control the charge and discharge
process.
Concerning the type of connectors, the IEC 62196-2 proposes
different types based on other standards the SAE J1772-2009 (Type
1) or the VDE-AR-E 2623-2-2 (Type 2). The type 3 is also xed, but
is not yet completely dened. Concerning the fast charging, the
62196-3 denes the DC fast charging plugs to be used in Mode 4,
namely the CHAdeMO and the combined charging systems.
In U.S. the charging standards are dened by the Society of
Automotive Engineers (SAE) in SAEJ1772. The recognized types of
plug are similar to IEC 62196, but SAE has selected the J1772
combo plug as the standard. This plug allows both AC and DC
charging using the same plug. The SAEJ17722011 denes six
charging levels, three in AC and three in DC [21]. The AC Level 1 is
practically applied at home environments while the AC Level 2 is
suitable for public and commercial areas. The DC-fast charging
levels (DC Level 13) are more adapted to public use.
In China, the EVs charging standards are dened in GB/T
202342011. According to this standard, EVs charging can be made
in AC using the Mode 2 and 3 dened in IEC 62196. However, this
use is limited to a single phase system allowing a maximum current of 16 A. The DC fast charging is also available considering a
specic connector that allows a maximum voltage of 750V and
125/250A [60]. The fast charging uses the CATARC (China Automotive Technology and Research Center) protocol. In Japan, the
CHAdeMO specications are used as a standard [61].
3.3. Communications standards
In order to illustrate how the charging and discharging schedule is implemented, this subsection discusses the relevant communication standards for integrating EVs into power grids. It is
noted that the purpose of this section is to provide the relevant/
widely used communication standard that can support EV smart
charging rather than comparing the various communication
standards. The IEC standards illustrated in Fig. 2 is recommended
in [62]. The objectives of the study [62] is to realize a standardized
communication interface between the vehicle and the grid. The
standardization will make it possible for EV users to have easy
access to EV charging equipment (EVSE) and related services
throughout Europe. EVSE refers to all devices installed for delivering power from the electrical supply point to the EV and this
charging equipment will support smart charging functions. The
decision of the charging can be made on the EV level or on the FOs
level. The IEC 15118 is the most recommended communication
standard [62], and is demonstrated in details in [63], by the
sequence diagram of a charging process between the EVSE and
the EVs.
The IEC 61850 is recommended for the communication
between the EVSE and the FOs [63]. In [64], the use of the smart
metering infrastructure to transmit the information concerning
the EVs charging state using the multiple access control protocol is
1214
Fig. 2. Relevant ICT standards support the EV smart charging in the context of smart grids.
proposed. According to [65], the protocol SAE J2836/3TM, published in January 2013, allows the coordination of both distributed
energy resources and electric vehicles including the V2G capability
(control the charge and discharge process). The information
exchange with the EV was derived from the IEC/TR 61850-90-8
from February 2013. In [66], two other promising protocols to be
used in the communication between charging stations and the
system operators and/or aggregators, namely Open Charge Point
Protocol (OCPP) and Hubject's Open Intercharge Protocol (OICP)
are analyzed.
http://www.powermatcher.net.
Centralized
control
Transactive
control
Price control
Computational
requirements
Features
1215
One/Two way
High
Two way
Relative high
One way
Low
Communication
form and cost
1216
Subject to
8
As P b Z bs
>
>
>
>
< Ag P b Z bg
Ab P b Z bb
>
>
>
>
: b r P r bu
l
b
Fig. 6. Information ow between the FO and the EVs in a price based control
strategy.
min
t max X
lines
X
t 1l1
Rl :I 2l;t
Subject to
8
8 t; 8 n A nodes :0 r P n;t r P max
>
>
>
>
t max
<
X
8 n A nodes :
P n;t :t:xn C max
>
>
t1
>
>
:
xn A 0; 1
where uk is the control variable for time k that denotes a dimensionless and discrete representation of P k . P k corresponds to the
purchased power. The total cost of a whole charging sequence, J is
calculated as:
J 0 x0 J N xN
N
1
X
lk xk ; uk ; k
k1
1217
In that case, the Pmin has been dened as zero for the charging
and discharging process, but in this new formulation the charging
power Pc,t and the discharging power Pd,t can be 0 or assume a
value between the minimum and maximum. The binary variables
xc,t and xd,t have been introduced to determine when (time period)
the EVs are charging or discharging.
5.5. Mixed-integer non-linear programming
The mixed-integer non-linear programming (MINLP) is used in
the EVs scheduling mainly when other distributed resources are
considered due to their non-linear operation characteristics, and
when it is necessary to introduce the network technical constraints such as the lines thermal limits and the bus voltage
operation boundaries [86]. The objective functions can be technical or economic taking into account the power losses [87], the
operation costs [82], the load balancing [37], the voltages proles
[88], the greenhouse gas emissions [48] among others.
1218
j A Li
Q Git Q Dit V it
X
jAL
V jt Gij sin ijt Bij cos ijt Bii
where V it and it are the voltage magnitude and angle in each bus i
and in each period t, respectively. The lines characteristics are
represented by the variables Bij , and Gij (imaginary and real parts of
the admittance matrix corresponding to the i row and j column).
Finally, generated power from different resources are represented by
P Git and Q Git and the consumption power including loads and
storage systems charge (EVs and storage systems) are represented by
the P Dit and Q Dit . Additionally, the MINLP formulation includes the
same integer variables that are used in MILP.
In general, the use of MINLP is suitable for problems with small
number of EVs and other resources (small number of variables).
For a large number of resources, the computational requirements
and the execution time turn the use of this methodology inadequate for real application. But, the results obtained are more exact
and can takes into account more realistic models of the devices
connected to the network. This means that the MINLP can be used
in small cases to determine the errors of each approximation made
by other techniques [82] or in real applications when a problem is
restrict to a small area with few number of resources.
E v^ n can be estimated by generating M independent samples
i;j of the random input variable each of size N. The M optimization
problems are performed where the nonlinear power ow equations are solved by using the backward-forward sweep method
[91]. The optimal values of M samples constitute a normal distribution:
(
)
N
1X
i;j
^v jn min f^ N P n;t
gP n;t ; ; j 1; ; M:
12
Ni1
j
v^ n is the mean optimal value of the problem for each of the M
samples. LN;M is an unbiased estimator of Ev^ n . Simulations indicate that in this type of problem, the lower bound converges to the
real optimal value when N is sufciently high:
LN;M
M
1 X
j
v^
Mj1 N
13
11
In the rst stage, the dispatch cost and the regulation capacity
cost are considered. In the second stage, the worst case
performance-based regulation cost is used. The variables P and R
represent the energy dispatch decision and the regulation amount
for automatic generation control (AGC) units presented in the rst
stage, respectively. The variable p(d) represents the secondstage energy output from the regulation unit, which is adjustable
to each demand realization d d, for d A D, in which D represents uncertainty set. In the economic dispatch problem, the
electric vehicles can be considered as a load in charge mode and as
generator in discharge mode in the objective functions. The
technical aspects of the EVs are considered in the constraints that
subject to the objective functions.
5.8. Heuristic and meta-heuristic algorithms
One of the main problems of the EVs scheduling is its interdependency between periods. For instance, the energy stored in
the battery in a certain period depends on the decisions applied in
previous periods, such as charging/discharging power or energy
spent in trips, and also on the decision applied in that period.
Besides, a large-scale penetration of EVs can increase signicantly
the number of variables and consequently the execution time to
nd an optimal solution in an optimization problem. Considering
the complexity, e.g., in [86], a deterministic technique (mixedinteger non-linear programming) is implemented to solve the dayahead resource scheduling considering a penetration of 2000 EVs
and this technique took around 28 h to obtain the optimal solution. In this kind of problems, it is easily that a large number of
variables and constraints is reached, such as 100,000 problem
variables in [100]. This problem can be classied as a NP-hard
problem [101]. The NP-hard problems include, no polynomial time
algorithms, i.e. algorithms that need exponential computation
time in the worst case to obtain the optimum solution [102].
The heuristics optimization algorithms intend to solve these
problems obtaining a solution near to the optimal (not necessary
the optimum) one in a convenient execution time. For example, in
the previously mentioned study [86], the proposed meta-heuristic
took around 18 min to nd a solution with a difference close to the
optimal one in 0.97%. In order to adapt the heuristics to real problems and to improve their performance, it is necessary to include
some additional algorithms in some steps of the heuristics process
that results in meta-heuristic algorithms.
An extended overview of the population-based meta-heuristic
techniques is presented in [103]. Regarding the use of metaheuristic in power systems, a review is presented in [104].
Recently, meta-heuristic methods applied to renewable energy is
reviewed in [102]. Thus, the following paragraphs intend to
overview the application of some meta-heuristic techniques in the
EVs charge and discharge scheduling.
In [100], a variant of PSO is proposed for solving the EVs charge
and discharge process that minimizes the operation cost of electric
vehicles. The main feature of the proposed method is the inclusion
of a process to change the particles velocity in each iteration.
Articial bee colony (ABC) is used in [105] to optimize the management of distributed energy resources, including distributed
generation, demand response and EVs. An ABC-based multiobjective algorithm is proposed in [106] to optimally determine
the number, locations and sizes of the distributed generation and
parking lots.
Genetic Algorithms (GA) is used in [107] to optimize the EVs
charge scheduling that aims at minimize the operation costs while
taking into account the network supply constraints. In [108], a
modied GA algorithm is used to optimize the PHEVs charge
patterns that consider multi objectives: 1) the total cost of fuel and
electricity and 2) the total battery health degradation over a 24 h
1219
period. The papers concluded that these two objectives are conicting which resulting in a Pareto front optimal charge pattern.
Simulated annealing (SA) is proposed in [109] to schedule the
EVs charging and in [110] to schedule the charge and discharge
process, including the AC power ow constraints. One of the main
disadvantages of SA is its dependency of the initial solution.
Hence, in order to overcome this characteristic, a hybrid approach
is proposed in [86] using an ant-colony optimization (ACO) algorithm to determine the initial solution that will be used by SA. The
same authors [111] proposed two heuristics to obtain a good initial
solution for the EVs and distributed generation, respectively. These
heuristics presented an initial solution worse than the ACO in [86],
however shown a much lower execution time than the ACO
algorithm. Beyond the mentioned use of ACO to provide the initial
solution to SA proposed in [86], a hybrid PSO and ACO algorithm is
proposed in [112] to deal with short-term unit commitment problem with EVs considering different reliability limits.
1220
Table 5
General comparison between the presented algorithms.
Execution time
Certainty of performance
Applicability in general
Applications to EV charging
Ref. [78] showed that the calculation time using the Ref. [78] showed that the difference between using
QP is 819 times than the one using LP for a eet of LP and QP is minor. Therefore, the benet of using
the QP does not justify the increase in computation
50 vehicles.
time.
Ref. [80] indicated that the computational time for Ref. [80] showed that the difference between the
DP is slower compared to QP.
charging prole of using QP and DP is negligible,
although the QP gave more accurate results.
Control algorithms
n
xnt n unt ; t
0; ; T 1
16
where xn is the state of charge of EV n , n and n means the charging efciency and battery size of EV n , and un represents the local
control variable. The purpose of the study is to nd the set of
feasible full charging controls, which are described as follows:
17
n un un0 ; ; unT 1 ;s:t:unt Z 0; xnT 1
1221
where the nal constraint on xnT requires that all EVs are fully
charged by the end of the interval. The cost function of agent n,
denoted by J n u is used as criteria and specied as:
J n u
TX
1n
2 o
pr t unt unt avg ut
18
t0
i dem i dem
P r j P r
o
f d r ; 8 r A 0; ; rmax
20
[121]
[120]
Applications to EV charging
Control/computing algorithms
Key features
References Modeling methods
Table 6
General comparisons among the presented studies.
1222
1223
Table 7
General comparisons among the presented studies.
References Modeling methods
Key features
[74]
[123]
[75]
[124]
Price signal
format
Time-of-use
tariff.
A social experiment is suggested to estimate the price response of Time-of-use
EVs' charging.
tariff.
1) Parametric utility model.
Compared to [74], utility functions are developed for the three
Dynamic price
2) Stochastic expectation model.
types loads: critical, interruptible and deferrable load.
signal.
Dynamic price
Distribution locational marginal pricing It is discussed that compared with linear programming based
signal.
method via QP.
DLMP formulation [128], QP based formulation ensures convergence to the DLMP.
time-of-use rates are used and each rate has three pricing periods:
peak, mid-peak, and off-peak. In the simulation study, the authors
developed a conditional logit model to get the choice probabilities
of EV charging in presences of different time-of-use rates. The
simulation results shown that a value of 0.80 will be needed to
effectively eliminate peak time charging introduced by electric
vehicles' charging and a value of 0.25 will be needed to eliminate half of the normal peak time charging load, after running
simulations with a wide range of price elasticity.
Yu et al. [75] investigated the price elasticity of electricity
consumers and these are also the important aspects in price
control method. In [75], the marginal utility function of loads is
realized by the following parametric stochastic process
8
< t ; r t r ;
rt
21
otherwise:
: 0;
where ; ; ; are random variables that describes the different
characteristics of utility function, stands for the time slot that a
task is initially requested, which also reects the task distribution,
is the initial marginal utility, which stands for the magnitude of
the marginal utility, is the tolerable delay, which determines the
maximum delay that a user can tolerate to nish a task, means
the utility decay rate, which represents the cost of inconvenience
by the delay.
Using the model, the scheduling of each individual task is now a
random event whose probability distribution is controlled by the
stochastic process rt. The aggregate demand curve can be estimated through expectation with respect to the distribution of rt.
Note that some assumptions have been made in [75], such as the
time period of the scheduling is divided into T time slots, there are
total M individual tasks m: m1,, M of different appliances that
are to be initialized by all the users within the scheduling period,
and each task will consume xm kW h. Furthermore, it is assumed
that each task can be completed within one time slot; therefore,
tasks that have duration longer than one time slot will be decomposed into multiple tasks that are considered independently.
Huang et al. [124] proposed a distribution locational marginal
pricing (DLMP) method through quadratic programming designed
to alleviate congestions that might occur in a distribution network
with high penetration of exible demands such as electric vehicles
and heat pumps. In the DLMP method, the DSO calculates dynamic
tariffs and sends them to the eet operators. The eet operators
make the optimal energy schedules for the exible demands. It is
proven in the study that the DLMP through quadratic programming modeling can ensure a unique marginal price for the FOs.
Instead, DLMP through linear programing modeling [128] might
bring multiple solutions issue of the eet operator optimization
that may cause the congestion management by DLMP to fail. The
case studies showed that DLMP helps alleviating congestion problem. It is also shown that the linear programming based modeling can lead the failure of the DLMP.
Applications to EV charging
In summary, studies [74,75,123,124] suggested that the electricity price can be properly designed to reduce the peak demand
as EVs penetrate in the vehicle market. However, it is also noted in
[74,75,123] that to which extent the properly designed price signal
could assist in maintaining grid reliability will remain open until
empirically tested EV owner's price responsiveness through
experiment pilots are known. When investigating EV owner's
price responsive behavior, decomposition methods [26] widely
used in energy economics eld can be applied to nd the relation
between EV's charging response and electricity price, state of
charge, customer driving habits and reasons for buying the EV etc.
Additionally, Table 7 depicts the information of the presented
algorithms in terms of modeling methods, key features, price
signal format and applications to EV charging. The presented
algorithms are also summarized and compared in this table.
1224
control and price control are discussed, which gives a basis for
comparing available methods for future developments;
Details of the modeling method and algorithms of each control
strategy are illustrated by showing the key formulas and compared in term of their performance, calculation time, communication and computation requirements etc.
It should be observed and emphasized that the above discussion did not consider the real time operations, i.e., there is no
continuous monitoring and assessment of the state of electric
vehicle battery dynamics and therefore lack of the appropriate
response in abnormal situations. This means that real time control
method considering the dynamic behavior of EV eet and power
systems should be developed as well.
References
[1] European Commission, Energy Roadmap 2050; 2011.
[2] Wang Z, Yang L. Delinking indicators on regional industry development and
carbon emissions: BeijingTianjinHebei economic band case. Ecol Indic
2015;48:418.
[3] Wang Z, Yin F, Zhang Y, Zhang X. An empirical research on the inuencing
factors of regional CO2 emissions: evidence from Beijing city, China. Appl
Energy 2012;100:27784.
[4] Blesl M, Das A, Fahl U, Remme U. Role of energy efciency standards in
reducing CO2 emissions in Germany: an assessment with TIMES. Energy
Policy 2007;35(2):77285.
[5] Wang Z, Wang C, Yin J. Strategies for addressing climate change on the
industrial level: affecting factors to CO2 emissions of energy-intensive
industries in China. Nat Hazards 2014;75(S2):S30317.
[6] Zhang B, Wang Z. Inter-rm collaborations on carbon emission reduction
within industrial chains in China: practices, drivers and effects on rms'
performances. Energy Econ 2014;42:11531.
[7] European Environment Agency, Greenhouse gas emission trends, 2013.
[8] Ellabban O, Abu-Rub H, Blaabjerg F. Renewable energy resources: current
status, future prospects and their enabling technology. Renew Sustain
Energy Rev 2014;39:74864.
[9] Paliwal P, Patidar NP, Nema RK. Planning of grid integrated distributed
generators: a review of technology, objectives and techniques. Renew Sustain Energy Rev 2014;40:55770.
[10] Sun Z, Li K, Yang Z, Niu Q, Foley A. Impact of electric vehicles on a carbon
constrained power systema post 2020 case study. J Power Energy Eng
2015;03(04):11422.
[11] Fadaeenejad M, Saberian AM, Fadaee M, Radzi MAM, Hizam H, AbKadir MZA.
The present and future of smart power grid in developing countries. Renew
Sustain Energy Rev 2014;29:82834.
[12] Gouveia C, Rua D, Soares FJ, Moreira C, Matos PG, Lopes JAP. Development
and implementation of Portuguese smart distribution system. Electr Power
Syst Res 2015;120:15062.
[13] Parks K, Denholm P, Markel AJ. Costs and emissions associated with plug-in
hybrid electric vehicle charging in the Xcel Energy Colorado service territory.
CO: National Renewable Energy Laboratory Golden; 2007.
[45] Bellekom S, Benders R, Pelgrm S, Moll H. Electric cars and wind energy: two
problems, one solution? A study to combine wind energy and electric cars in
2020 in The Netherlands Energy 2012;45(1):85966.
[46] Dallinger D, Wietschel M. Grid integration of intermittent renewable energy
sources using price-responsive plug-in electric vehicles. Renew Sustain
Energy Rev 2012;16(5):337082.
[47] Saber AY, Venayagamoorthy GK. Plug-in vehicles and renewable energy
sources for cost and emission reductions. IEEE Trans Ind Electron 2011;58
(4):122938.
[48] Saber AY, Venayagamoorthy GK. Resource scheduling under uncertainty in a
smart grid with renewables and plug-in vehicles. IEEE Syst J 2012;6(1):103
9.
[49] Schuller A, Flath CM, Gottwalt S. Quantifying load exibility of electric
vehicles for renewable energy integration. Appl Energy 2015;151:33544.
[50] Birnie DP. Solar-to-vehicle (S2V) systems for powering commuters of the
future. J Power Sources 2009;186(2):53942.
[51] Zhang Q, Tezuka T, Ishihara KN, Mclellan BC. Integration of PV power into
future low-carbon smart electricity systems with EV and HP in Kansai Area,
Japan. Renew Energy 2012;44:99108.
[52] Hu J, You S, Lind M, stergaard J. Coordinated charging of electric vehicle for
congestion prevention in the distribution grid. IEEE Trans Smart Grids
2014;5(2):70311.
[53] Kristoffersen TK, Capion K, Meibom P. Optimal charging of electric drive
vehicles in a market environment. Appl Energy 2010;88:19408.
[54] Al-Awami AT, Sortomme E. Coordinating vehicle-to-grid services with
energy trading. IEEE Trans Smart Grid 2012;3(1):45362.
[55] Johannsson H, Hansen H, Hansen LH, Holm-Hansen H-H, Cajar P, Bindner
HW, Samuelsson O. Coordination of system needs and provision of services.
In: Proceedings of the 22nd international conference and exhibition on
electricity distribution (CIRED 2013); 2013. p. 14.
[56] Sbordone D, Bertini I, Di Pietra B, Falvo MC, Genovese A, Martirano L. EV fast
charging stations and energy storage technologies: a real implementation in
the smart micro grid paradigm. Electr Power Syst Res 2014.
[57] Guzzella L, Sciarretta A. Vehicle propulsion systems: introduction to modeling and optimization. Berlin, Heidelberg: Springer; 2007.
[58] Zhang C, Li K, Mcloone S, Yang Z. Battery modelling methods for electric
vehicles a review. In: Proceedings of the European control conference
(ECC); 2014. p. 267378.
[59] Blockx P, Van Den Bossche P, Omar N, Van Mierlo J, Rotthier B, Cappelle J.
Beyond the plug and socket: towards safe standardized charging infrastructures. In: Proceedings of the world electric vehicle symposium and
exhibition (EVS27); 2013. p. 110.
[60] Tang WW, Wu YM, Qin J. Comparative study on electric vehicle charging
standards at home and abroad. Adv Mater Res 2013;608609:15539.
[61] Bakker S. Standardization of EV recharging infrastructures; 2013.
[62] Ka bisch S, Schmitt A, Winter M, Heuer J. Interconnections and communications of electric vehicles and smart grids. In: Proceedings of the rst IEEE
international conference on smart grid communications (SmartGridComm),
IEEE; 2010. p. 161166.
[63] Hansch K, Pelzer A, Komarnicki P, Groning S, Schmutzler J, Wietfeld C, Heuer
J, Muller R. An ISO/IEC 15118 conformance testing system architecture. In:
Proceedings of the IEEE PES general meeting|conference & exposition; 2014.
p. 15.
[64] Yang Y, Roy S. Grouping-based MAC protocols for EV charging data transmission in smart metering network. IEEE J Sel Areas Commun 2014;32
(7):132843.
[65] American National Standards Institute. Standardization roadmap for electric
vehicles; 2013.
[66] Mltin M, Gitte C, Schmeck H. Smart grid-ready communication protocols
and services for a customer-friendly electromobility experience. In: GI-Jahrestagung; 2013. p. 147084.
[67] Shahidinejad S, Bibeau E, Filizadeh S. Statistical development of a duty cycle
for plug-in vehicles in a North American urban setting using eet information. IEEE Trans Veh Technol 2010;59(8):37109.
[68] Soares J, Canizes B, Lobo C, Vale Z, Morais H. Electric vehicle scenario
simulator tool for smart grid operators. Energies 2012;5(12):188199.
[69] Azadfar E, Sreeram V, Harries D. The investigation of the major factors
inuencing plug-in electric vehicle driving patterns and charging behaviour.
Renew Sustain Energy Rev 2015;42:106576.
[70] Rolim CC, Gonalves GN, Farias TL, Rodrigues . Impacts of electric vehicle
adoption on driver behavior and environmental performance. Procedia - Soc
Behav Sci 2012;54:70615.
[71] Ipakchi A. Demand side and distributed resource managementa transactive
solution. In: Proceedings of the power and energy society general meeting,
IEEE; 2011. p. 18.
[72] Melton RB. GridWise transactive energy framework (DRAFT Version). Pacic
Northwest National Laboratory (PNNL), Richland, WA (US); 2013.
[73] Kok JK, Warmer CJ, Kamphuis IG. PowerMatcher: multiagent control in the
electricity infrastructure. In: Proceedings of the fourth international joint
conference on autonomous agents and multiagent systems; 2005. p. 7582.
[74] Shao S, Zhang T, Pipattanasomporn M, Rahman S. Impact of TOU rates on
distribution load shapes in a smart grid with PHEV penetration. In: Proceeedings of the IEEE PES transmission and distribution conference and
exposition: smart solutions for a changing world; 2010.
[75] Yu R, Yang W, Rahardja S. A statistical demand-price model with its application in optimal real-time price. IEEE Trans Smart Grid 2012;3(4):173442.
1225
1226
[118]
[119]
[120]
[121]
[122]
[123]
[124]
[125]
[126]
[127]
[128]
IEEE Vehicle Power and Propulsion Conference (VPPC), September 3-5, 2008, Harbin, China
attention.
The co-existence of several control subsystems can
cause two main problems. First, the design of software
and hardware will become more complicated due to the
dramatically increased number of sensors and signal
cables. Second, because of the possible function
overlapping among these systems, their conflicts of
control objectives and actions will inevitably occur. If
without coordination, the dynamic performance of these
systems could be worse than that of individual system or
even worse than a passive system without any active
control. For example, ABS and ESP are both based on the
tire slip ratio control, however the primary objective of
ABS is to maintain the slip ratio around the optimal value
(normally corresponding to peak friction coefficient),
while the objective of ESP is to control wheel slip ratios
properly for vehicle stability improvement via braking the
properly-selected wheels.
Therefore, the two key problems of integrated vehicle
dynamics control to be solved are: 1) to avoid the
conflicts and interventions among different subsystems;
2) to exploit the potentials of each subsystems by
communication and coordination among the sub-systems
since different system has different action domain.
In practice, vehicle electric control systems could cover
much broader domain, including body electric control
systems, but this paper is mainly to review the Integrated
Vehicle Dynamics Control (IVDC), also called
Integrated Chassis Control based on many literatures
published within recent 20 years. Although many relevant
researches have been published recently, the definition of
IVDC is still not very definitive. However, its attributes is
obvious, i.e. it needs to coordinate two or more
subsystems systematically according to control objectives
and actions in terms of both software and hardware,
rather than simply put the subsystems altogether.
Besides the obvious benefits of IVDC, including
reduction of number of sensors and actuators, less
mounting space and weight, etc., IVDC can further
exploit the potential benefits of the control systems and
thus enhance overall vehicle dynamic performances. By
integrating different subsystems, the comprehensive
performance of IVDC can be much better with the
potential benefits as shown in Figure 1, in which the
sphere diameter can be maximized by system
coordination. Hence, overall vehicle dynamics
performances, in directions of longitudinal, lateral and
vertical, can be improved.
IEEE Vehicle Power and Propulsion Conference (VPPC), September 3-5, 2008, Harbin, China
Fig 1.
IEEE Vehicle Power and Propulsion Conference (VPPC), September 3-5, 2008, Harbin, China
IEEE Vehicle Power and Propulsion Conference (VPPC), September 3-5, 2008, Harbin, China
X = = A + B r + H f
r
M za
r
Other researchers linearized the non-linear vehicle
model under some certain working conditions, then
designed the controller based on linear model by using
robust theory [20, 24]. This method dealt with the
non-linear factors as uncertainties before modeling by
presetting the control gain to counteract the undesired
IEEE Vehicle Power and Propulsion Conference (VPPC), September 3-5, 2008, Harbin, China
REFERENCES
[1] Tokuda T. Cars in the 90s as a humanware. SAE Paper 885049,
1988
[2] Kizu R., Harada H., Minabe H. Electronic control of car chassis
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
IEEE Vehicle Power and Propulsion Conference (VPPC), September 3-5, 2008, Harbin, China
Taiwan, 2006
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of
Nature Science Foundation of China (Grant No. 50575141)
and the Opening Fund (Grant No. 20071111) by Key State
Laboratory of Automobile Dynamic Simulation, Jilin
University, www.ascl.jlu.edu.cn .
Jurnal
Teknologi
Full paper
Faculty of Mechanical Engineering, Universiti Teknologi Malaysia, 81310 UTM Johor Bahru, Johor, Malaysia
Technology Center, Transportation Research Alliance, Universiti Teknologi Malaysia, 81310 UTM Johor Bahru, Johor, Malaysia
bMarine
Article history
Abstract
This paper presents a review of research work done on various aspects of control system approaches of
unmanned surface vehicle (USV) in order to improve the course keeping performance. Various methods
have been used to produce a course keeping control system for manoeuvring system of USV. However,
the review reveals that the adaptive backstepping control system is a powerful tool for the design of
controllers for nonlinear systems or transformable to form a tight feedback parameter. It is very suitable
for the automated control system of USV in relative motion that involves the disturbances from waves
and wind. Fuzzy logic control also had been suggested as an alternative approach for complex systems
with uncertain dynamics and those with nonlinearities. This method does not rely on the mathematical
models, but the heuristic approach. Further studies may be conducted to combine the control method
approach mentioned above to develop a real time system with robust control laws to the motions of a
USV in waves, usually at a specific speed, including station keeping or heading in sinusoidal and irregular
waves.
Graphical abstract
1.0 INTRODUCTION
Unmanned surface vehicle (USV) fills an increasingly important
knowledge in the marine engineering study for robotic vehicles.
Research and development in USV is a specialization in the realm
of marine robotics which consists of guidance, navigation, and
control as basic approach for controlling surface vehicles [1].
USVs are also called Autonomous Surface Crafts (ASCs) or
Autonomous Surface Vehicles (or Vessels) (ASVs) [2-6], is a
type of vessel that runs on the surface of water without an
operator on the platform and allows new modes of operations [2].
On the design of the USV, there are mostly unique small
designs with special purpose craft and not standardized for
modularity [6]. This design has limited endurance, payload, and
seakeeping ability. In the current design of USV, most small
vehicles such as boats have been adapted from ships manned
originally designed to accommodate a human occupant [7]. This
design provides moderate speed with long endurance and
improved capabilities of payload and seakeeping to support
special missions.
Thus, USVs can be used for a variety of missions that will be
safer and cheaper than human, such as marine environment
monitoring, hydrology survey, target object searching,
bathymetric mapping, defence, general robotic researches, and
scientific studies. [6, 7]. Some of the missions above need
advanced control system to manoeuvre the USV. Therefore, this
paper is intended to present a literature review of research work
12
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
13
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
14
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
15
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
16
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
17
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
18
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
Table 1 Researches on course keeping control system approaches for USV
6.0 CONCLUSION
In this study, the literature related to course keeping control
system for USV has been reviewed. Simple classifications of
different course keeping control approaches have been defined.
The controllers were based on a two dimensional planar model of
the system that included surge, sway, and yaw dynamics. The
review reveals the significant progress made during the last two
decades in the development of USV and the course keeping
control system. However, the review also reveals that adaptive
backstepping is a powerful tool for the design of controllers for
nonlinear systems or transformable to form tight feedback
parameters. This method is suitable for the automated control
system to USV in relative motion that involves a disturbance from
waves and wind. Fuzzy logic control also has been suggested as
an alternative approach for complex systems with uncertain
dynamics and those with nonlinearities. This type of method does
not rely on mathematical models but the heuristic approach.
19
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
20
M. N. Azzeri, F. A. Adnan & M. Z. Md. Zain / Jurnal Teknologi (Sciences & Engineering) 74:5 (2015), 1120
[48] Strand, J. P., Ezal, K., Fossen, T. I. 1998. Nonlinear Control of Ships: A
Locally Optimal Design. In: Preprints of The IFAC NOLCOS98,
Enschede, Netherlands. 732738.
[49] Nassim, K., Nabil, G. C. 2013. A Self-Tuning Guidance and Control
System for Marine Surface Vessels. J. Nonlinear Dynamics. 73(12):
897906.
[50] Fossen, T. I. 1994. Guidance and Control of Ocean Vehicles. John Wiley
and Sons, Chichester, England.
[51] Zoran, Vc, Bruno, Bc. 2001. Guidance and Control Systems for Marine
Vehicles. in The Ocean Engineering Handbook, F. El-Hawary, Ed. 1st
ed. Boca Raton CRC Press. 2527.
[52] Holzher, T., Schultze, R. 1996. Operating Experience with a HighPrecision Track Controller for Commercial Ships. Control Engineering
Practice. 4(3): 343350.
[53] Minorsky, N. 1992. Directional Stability of Automatically Steered
Bodies. J. American Society for Naval Engineers. 34(2): 280309.
[54] Van Amerongen, J. 1982. Adaptive Steering of Ships-A Model Reference
Approach to Improved Maneuvering and Economical Course Keeping,
Delft University of Technology, Ph.D. Thesis, Delft, Netherlands.
[55] Hearn, G. E., Zhang, Y., Sen, P. 1997. Alternative Designs of Neural
Network Based Autopilots: A Comparative Study. In: Proc. of 4th IFAC
Conference on Maneuvering and Control of Marine CraftMCMC 97,
New York, USA.
[56] Kumar, V., Nakra, B.C., Mittal, A. P. 2011. A Review on Classical and
Fuzzy PID Controllers. Int. J. Intelligent Control and Systems. 16(3):
170181.
[57] Park, J-H, Shim, H-W, Jun, B-H, Kim, S-M, Lee, P-M, Lim Y-K. 2010.
A Model Estimation and Multi-Variable Control of an Unmanned
Surface Vehicle with Two Fixed Thrusters. In: OCEANS 2010 IEEE
Sydney. 15.
[58] Juntong, Q., Yan, P., He, W., Jianda, H. 2007. Design and Implement of
a Trimaran Unmanned Surface Vehicle System. In: International
Conference on Information Acquisition, 2007. ICIA '07, Seogwipo,
South Korea. 361365.
[59] Donald, E. K. 1998. Optimal Control Theory: An Introduction. PrenticeHall, New Jersey, U.S.A
[60] Sharma, S. K., Naeem, W., Sutton, R. 2012. An Autopilot Based on a
Local Control Network Design for an Unmanned Surface Vehicle. J.
Navigation. 65(2): 281301.
[61] Athans, M. 1971. The Role and Use of The Stochastic Linear-QuadraticGaussian Problem in Control System Design. J. Automatic Control,
IEEE. 16(6): 529552.
[62] Siramdasu, Y., Fahimi, F. 2012. Incorporating Input Saturation for
Underactuated Surface Vessel Trajectory Tracking Control. In: American
Control Conference (ACC), 2012, Montreal, Canada. 62036208.
[63] Sharma, S. K., Sutton, R., Motwani, A., Annamalai, A. 2013. Non-Linear
Control Algorithms for an Unmanned Surface Vehicle. Proceedings of
the Institution of Mechanical Engineers, Part M. Journal of Engineering
for the Maritime Environment 2008. 227(4): 110.
[64] Annamalai, A., Motwani, A., Sutton, R., Yang, C., Sharma, S.,
Culverhouse, P. 2013. Integrated Navigation and Control System for an
Uninhabited Surface Vehicle Based on Interval Kalman Filtering and
Model Predictive Control. In: IET Conference on Control and
Automation 2013. 16.
[65] Astrom, K. J., Wittenmark, B. 1995. Adaptive Control. Addison-Wesley,
Reading, Massachusetts.
[66] Junfang, L., Tieshan, L., Zhongzhou, F., Renxiang, B., Qiang, L.,
Jiangqiang, H. 2011. Robust Adaptive Backstepping Design for CourseKeeping Control of Ship with Parameter Uncertainty and Input
Saturation. In: International Conference of Soft Computing and Pattern
Recognition (SoCPaR), Dalian, China. 6367.
49
24
2013
12
Vol.49 No.24
Dec.
2013
DOI
10.3901/JME.2013.24.108
NVH
1, 2 1
201804
2. 153-8505 )
(1.
NVH
(Noise, vibration and harshness, NVH)
NVH
U46
1, 2
LUAN Wenbo
present overseas and domestic researches are concluded in this paper. The prime research contents and routes of vehicle driveline
NVH problems are also introduced. The discussions are unfolded from two aspects of simulation and experiment. Aiming at dynamic
research of NVH problems related to automotive driveline, the characteristics of four mechanical models which are systematic
dynamics model, multi-body dynamics model, structural dynamics model, systematic and structural dynamics hybrid model are
analyzed in detail, coordinates and reference systems used in modeling and the main simulation quantitative and qualitative research
methods are also reviewed in simulation. Meanwhile, research contents, signal processing methods, subjective and objective
assessment are presented briefly in experiment research aspect.
Key wordsDriveline of automotive Vibration Noise Harshness Dynamic model
(Noisevibration and
harshnessNVH)NVH
NVH
(51175379)20130715 20131128
NVH NVH
[1]
2013 12
NVH
109
NVH
NVH
NVH
NVH
NVH
NVH
NVH NVH
(1)
(2)
(3) (
)
NVH
NVH
NVH
NVH
NVH
1 (FF)
1
NVH
[2]
[3]
110
[5]
(Automated mechanical transmission
AMT) 3
[6](Torque converterTC)
15
[9]
[10]
4
NVH
[4]
[7]
[8]
[11]
[12]
[13]
[14]
49 24
[15]
[16]
(50500 Hz)
[19]
()(
)
[20]
NVH
NVH
[17]
[18]
NVH
NVH
( 2)
2013 12
NVH
NVH
(2)
(3)
(4)
(1)
NVH
111
NVH
3 NVH
NVH
()()
()
()--
()
NVH
3.1
NVH
NVH
(
)()
[21]
112
NVH
3.2
3.3
3.4
49 24
NVH
()
(
)
3.5
NVH
NVH
NVH
(1) ()
()
-
adams
-
-
4.1
[22]
2013 12
NVH
NVH
(2)
(3)
()
(
)
-
[23]
[24]
[25]
4.2
[25]
NVH
NVH
113
(
)
NVH
NVH
NVH
NVH
Gabor
46
[26]
[14]
114
NVH
[27]
[1]
[2]
10
NVH
(1) NVH
NVH
(2) NVH
(3)
NVH
NVH
[3]
49 24
[4]
[5]
[6]
, , .
[J]. , 2000, 21(3)22-26.
JIANG Guoping WANG Guolin, CHEN buda. Review
2013 12
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
115
NVH
USA SAE International, 2011
nonlinear vibration of gear pairs[J]. Journal of Sound and
Vibration, 2012, 331(15)3628-3648.
Warrendale, PA,
2011-01-1509.
WANG M Y, MANOJ R, ZHAO W. Gear rattle
modelling and analysis for automotive manual
transmissions[J]. Proceedings of the Institution of
Mechanical Engineers, Part DJournal of Automobile
Engineering, 2001, 215(2)241-258.
, , , .
[J]. , 1997, 16(3)391-396.
WANG Jianjun, LU Mingwan, ZHANG Xiong, et al. The
theory of rattle dynamics for automotive gear shift
system[J]. Mechanical Science Technology, 1997, 16(3)
391-396.
XIANG Dong, XIE Ning, SHEN Gang, et al. The
influence of rotation speed on gear rattle under
overdrive[C]// 3rd International Conference on Advanced
Engineering Materials and Technology, AEMT 2013.
Advanced Materials Research. SwitzerlandTrans. Tech.
Publications Ltd., 2013753-755, 960-967.
MEHMET B. Torsional vibration model based
optimization of gearbox geometric design parameters to
reduce rattle noise in an automotive transmission[J].
Mechanism and Machine Theory, 2010, 451583-1598.
, , . DMF-CS
[J]. , 2009, 20(15)
1787-1790.
LI Wei, LONG Yan, SHI Wenku. Analysis of isolation of
the torsional vibration of DMF-CS with centrifugal
pendulum-type absorber[J]. China Mechanical Engineering,
2009, 20(15)1787-1790.
, .
[J]. , 2011, 31(6)76-80.
SHI Wenku, MIN Delei. Study on vibration damping
characteristics of triple mass flywheel torsional damper[J].
Noise and Vibration Control, 2011, 31(6)76-80.
CURTIS S, PEARS J, PALMER D. An analytical method
to reduce gear whine noise, including validation with test
data[C]//Society of Automotive Engineers. SAE Technical
Papers. Warrendale, PA, USA SAE International, 2005
2005-01-1819.
.
[J]. , 2013, 32(11)95-99.
LUAN Wenbo, WU Guangqiang, XU Weiqing. Gearbox
whine noise analysis based on order tracking[J]. Journal
of Vibration and Shock, 2013, 32(11)95-99.
ERITENEL T, PARKER R G. Three-dimensional
116
, 2006.
1965
49 24
16
4 21 2
3 SAE
1 2 4
2 1
2
E-mailwuguangqiang@tongji.edu.cn
()1985
7 4
Ei 1
E-mailflashlwb_168@163.com
September, 2009
Vol. 2 No.3
conceptual framework of an agricultural vehicle autonomous guidance system, and then analyze its device characteristics.
This paper introduces navigation sensors, computational methods, navigation planners and steering controllers. Sensors
include global positioning systems (GPS), machine vision, dead-reckoning sensors, laser-based sensors, inertial sensors
and geomagnetic direction sensors. Computational methods for sensor information are used to extract features and fuse
data. Planners generate movement information to supply control algorithms.
into changes in position and direction.
proceeded to commercialization.
A number of prototype guidance systems have been developed but have not yet
GPS and machine vision fused together or one fused with another auxiliary
technology is becoming the trend development for agricultural vehicle guidance systems.
robotic technologies will augment the realization of agricultural vehicle automation in the future.
Keywords: agricultural vehicle, guidance, machine vision, GPS
DOI: 10.3965/j.issn.1934-6344.2009.03.001-016
Citation: Ming Li, Kenji Imou, Katsuhiro Wakabayashi, Shinya Yokoyama. Review of research on agricultural vehicle
autonomous guidance.
the 1920s[2,3].
Introduction
Researchers are
agricultural vehicles.
The
automated
vehicles
productivity,
agricultural
increased
potential
include
application
benefits
of
increased
accuracy,
and
Various
September, 2009
Vol. 2 No.3
[6,7]
Sensor
Performance results
Literature
Zhang[9,10]
Benson[11,12]
GPS
OConnor[13]
Tractor
Machine vision,
Mechanical sensor, GPS
Standard deviation of
position of 2.7 and 2.3 cm
strand
Machine vision
Tractor
Sgaard[17]
FOG, Ultrasonic
Doppler sensor
Imou[18]
Nagasaka
BRAIN, Japan
Tractor
Yukumoto[20]
Tractor
Noguchi [21,22]
Machine vision
Tractor
Accuracy of 2 cm
Billingsley[23]
[15,16]
[19]
In the
agricultural automation.
development.
September, 2009
Vol. 2 No.3
Figure 2 shows
factors,
Compared
with
these
complicating
For example,
[26]
Marchant
and Brivot
in real time (10 Hz) and noted that their technique was
as a stationary environment.
distinguished.
data points.
Navigation sensors
vehicle.
based sensing.
The effectiveness of
September, 2009
Vol. 2 No.3
determination
systems.
for
[28]
vision-based
[29]
guidance
also
They
binocular camera.
Han et al.
[25]
This
The guided
Okamoto et al.
[30]
developed
The method
vision.
Yukumoto et al.
[20]
Billingsley and
[23]
Schoenfisch
Their method,
The information
weeds.
The accuracy of
guidance information.
200 weeds/m2).
detection.
September, 2009
Vol. 2 No.3
3.2
technology
developments,
these
limitations
will
guidance sensors
[34-37]
systems.
for
farm
vehicles.
They found
Standard deviation of
technique.
tractors.
This integrated
September, 2009
Vol. 2 No.3
by the robot.
They
systems[45,46].
emitter.
straight line.
[19]
Nagasaka et al.
Root-mean-square deviation
from the desired straight path after correcting for the yaw
0.7 m/s.
low-speed motions.
3.4
Laser-based sensors
Laser-based sensors have a relatively longer range and
higher resolution.
[19]
However,
They
Dead-reckoning sensors
artificial landmarks.
With an average
However,
unbounded
accumulation
of
errors.
Specifically,
[50]
September, 2009
[51]
Vol. 2 No.3
distance of 20 m.
RTK GPS.
developed
rows.
The results
[14]
developed an autonomous
3.6
geomagnetic
direction
sensor
(GDS)
is
Tofael
[53]
compass[6].
other sensors.
Noguchi et al.[58] used a GDS to provide heading
[11]
high.
directional lines.
3.5
Inertial sensors
Inertial sensors take measurements of the internal
used
conditioner fan.
types
of inertial sensors
gyroscopes.
are accelerometers
Accelerometers
measure
and
acceleration
This includes
[54]
applications
September, 2009
Vol. 2 No.3
Ultrasonic
guidance.
[14]
(Subramanian et al.
[61]
).
Computational methods
shown in Figure 5.
Hough transform
The Hough transform technique can be used to isolate
The
4.2
Kalman filter
V.C. Hough.
still possible.
In a highly dynamic
Navigation
large.
are
from
structures
[15,26,27,52,56,62,63]
transform
using
the
Hough
[32]
used a stereovision-based
elevation
functions
map
using
Kalman
filters.
These integrated
root-mean-squared (RMS)
The algorithm
of stereo-image processing,
(INS)
consists
System
creation
and
navigation
point
September, 2009
Vol. 2 No.3
images.
Figure 6.
Navigation planners
Navigation planner plays an important role for
agricultural
vehicle
autonomous
control,
which
Row location
[28]
5.1
[69]
. Zhang et al.
[70]
Tracking methods
avoidance.
4.3
Other methods
[7]
However, if the
5.2
The
1) Dead reckoning
[25]
Han et al.
Freeland et al.[71]
10
September, 2009
Vol. 2 No.3
position information.
Nagasaka et al.
[72]
used rotary
2) Kinematic model
Unfortunately, pure
[73]
al.
Garaca-Prez et
sensors
wireless
electrohydraulic actuators.
variables
characterizing
the
sliding
effects
(laser
rangefinder,
communication
bumper,
system
inclinometers,
(WLAN)
and
Sensor-fusion algorithms
were
unknown
parameters
using
linearized
approximation.
3) Dynamic model
Dynamic models are fairly complex for agricultural
vehicle navigation, since describing all vehicle features
(e.g., inertia, sliding, springing) leads to very large,
[77-79]
4) Sensor fusion
Noguchi et al.
[58]
used an NN
September, 2009
Vol. 2 No.3
11
Zhu et al.
[80]
system delay.
speed sprayer.
NN vehicle model.
guidance results.
Ryerson
front wheels.
obstacles.
wheels.
Steering controller
Discussion
Since the time when the first driverless tractor
ground
parameters
Consequently, steering
conditions,
and
many
other
commercial products.
challenge.
automatic
advanced
system
steering
modifications
to
controllers
the
existing
include
hydraulic
Various
12
September, 2009
Vol. 2 No.3
there are others barriers that have not yet been resolved
be
[24]
Hague et al.
concluded that
fused
[19,61]
gyroscopy
with
another
[11]
, GDC
technology,
or laser radar
[14]
such
as
, to realize
drift errors, but the beacons take time to set up and may
The
Machine
An integrated
Electronics,
[14,25]
working environment.
dependability
varying conditions.
In urban environments,
of
mechanical
technology.
The
legal liability.
Although the
A sophisticated
machine.
As
machines
become
more
September, 2009
Vol. 2 No.3
13
in the future.
For agricultural
[References]
[1]
[2]
[3]
[4]
1939; 111(20): 7.
399.
[6]
vibrations.
enough
information
for
our
navigation
system.
[8]
Environment
technique
using
hyper
identification
317320.
[9]
working environment.
Conclusions
398.
systems.
14
September, 2009
Vol. 2 No.3
12.
307316. 2002.
63(1): 4956.
[34] Larsen W E, Nielsen G A, Tyler D A. Precision navigation
with GPS. Comput Electron Agric, 1994; 11(1): 8595.
[35] Gan-Mor S, Ronen B, Josef S, Bilanki Y. Guidance of
autonomous vehicle for greenhouse transportation, Acta
Hortic, 1997; 443: 99104.
[36] Bell T. Automatic tractor guidance using carrier-phase
differential GPS. Comput. Electron Agric, 2000; 25(1-2): 53
66.
[37] Yukumoto O, Matsuo Y, Noguchi N. Robotization of
agricultural vehicles description of the tilling robot. Journal
of Agricultural Engineering Research, 2000; 34: 107114.
[38] Abidine A Z, Heidman B C, Upadhyaya S K Hills D J.
Application of RTK GPS based auto-guidance system in
agricultural production. ASAE Paper No. O21152. ASAE,
St. Joseph, MI. 2002.
79.
September, 2009
Vol. 2 No.3
15
373. 2002.
multiresolution
Hough
transform.
Proceedings
of
51.
[57] Ishida M, Imou K, Okado A, Takenaga H, Honda Y, Itokawa
N, Shibuya Y. Autonomous tractor for forage production.
Journal of the Japanese Society of Agricultural Machinery,
1998; 60(2): 5966. (in Japanese).
[58] Noguchil N, Ishii K, Terao H. Development of an agricultural
mobile robot using a geomagnetic direction sensor and image
sensors. Journal of Agricultural Engineering Research, 1997;
67: 115.
[59] Harper N, McKerrow P. Recognising plants with ultrasonic
sensing
for
mobile
robot
navigation.
Robotics and
(March): 3540.
1999.
[67] Madhukar B R, Nayak R A, Ray J K, Shenoy M R. GPS/DR
integration using low-cost sensors. In Proc. of 12th
16
September, 2009
Vol. 2 No.3
101(4): 396410.
June 2004.
Zhang
Q, Reid
F,
Wu
D. Nonlinear
for
vehicle
using
landmarks
based
on
September, 2009
Vol. 2 No.3
17
Abstract: Efforts are underway to rehabilitate the irrigation districts, such as in the Rio Grande Basin in Texas.
Water
distribution network models are needed to help prioritize and analyze various rehabilitation options, as well as to
scientifically quantify irrigation water demands, usages, and losses, and to help manage gate automation. However,
commercially available software packages were limited in applications due to their high cost and operational difficulty.
This study aims to develop a modeling tool for modeling the water flow profile in irrigation distribution networks. The
goal of developing the modeling tool was to make the modeling process simple, fast, reliable and accurate. On the basis
of methodological study, the modeling tool has been developed for branching canal networks with the assumption of
steady gradually varied flow. The flow profile calculation of the tool was verified from a single channel with 1% root
mean squared error compared to the benchmark calculation and a branching network with 5% to 12% relative errors
compared to check point measurement along the network. The developed modeling tool will be able to play an
important role in water quantification for planning, analysis and development for modernization of irrigation systems.
Keywords: irrigation distribution network, modeling tool, flow profiling
DOI: 10.3965/j.issn.1934-6344.2009.03.017-026
Citation: Yanbo Huang, Guy Fipps. Developing a modeling tool for flow profiling in irrigation distribution networks.
Int J Agric & Biol Eng, 2009; 2(3): 1726.
Introduction
There
irrigation channels.
(http://www.sobek.nl/prod/index.html);
(Laboratoire
France;
dHydraulique
Parrish Engineering,
de
France,
Beaverton,
gates
(http://www.iihr.uiowa.edu/projects/
18
September, 2009
Vol. 2 No.3
Hydraulic
technique.
water
canalcad/index.html);
systems
Mike
11
(Danish
(http://www.dhisoftware.com/mike11);
model
Saint-Venant
for
canal
automation
design (http://canari.
montpellier.cemagref.fr/papers/sic30.pdf);
HEC-RAS
equations[6].
These
equations
are
of boundary conditions.
and
more
flow
equations.
(http://www.hec.usace.army.mil/software/hec-ras);
in
branching
canal
networks
(http://www.
engineering.usu.edu/bie/software/canalman.php).
easily
implemented
than
the
Saint-Venant
These
and CanalMan.
Model
Study area
gradients.
Australia.
[4]
Hidalgo et al.
downstream
This
pressure;
This
the project area, the elevations range from sea level in the
measured.
[5]
Islam et al.
presented a hydraulic
than 100 m.
the
implicit
four-point
Preissmann
for
September, 2009
Vol. 2 No.3
19
Figure 1 Service areas of the irrigation districts in Lower Rio Grande Valley of Texas
Computing methods
flow (USGVF).
3.1
momentum[6].
The
computation
and analysis.
step method.
were
classified
as[6]:
the
graphical-
20
September, 2009
Sf =
Because the
Basically a
This
4/3
(3)
n2 Q2
A R
Au + Ad
A=
2
Ru + Rd
R=
2
Vol. 2 No.3
yu +
2
u
2
d
v
v
+ S o x = yd +
+ S f x
2g
2g
(1)
If the stopping
irrigation
distribution
networks
are
yu = yd + (
and
2
d
2
u
v
v
) S o x + S f x
2g 2g
(2)
September, 2009
The same
Q = C fo LGo 2 g ( yu 0.5Go )
where
structure;
flow.
(5)
(4)
laterals.
21
Vol. 2 No.3
(6)
(7)
The
[9]
structure.
3) Find out the standard equation of a specific gate
structure for the determined flow regime: equation (4),
3.3
Gate calibration
Gate is the most popular structure for controlling
In general, four
(8)
22
September, 2009
log10(Cso)=a+b*log10(yd/Go )
C++
(9)
programming
Vol. 2 No.3
language
was
chosen
for
(Object-Oriented Programming).
Model validation
The water flow profiles in a single irrigation canal
formulated as:
/Q
p = Cq
(10)
where
(11)
respectively.
III. The performance of the calibration can be
residuals.
At
Two laterals are from the main canal through two sluice
implementation.
Figure 3
September, 2009
about 732 m.
Gate
Flow regime
Gate status
HS3
NA
Closed
NA
Q = Cnf Ly
comparison.
Cnfw = Cnf / 2 g
Open
Submerged
Orifice Flow
Submerged
Orifice Flow
-0.2917
Q = Cso Lhd 2 g ( yu yd )
Open
Q = Cso Lyd 2 g ( yu yd )
Open
C nfw =0.0469
HS6
1.5
u
Free Weir
Flow
scheme.
HS4
23
Vol. 2 No.3
and
gate
calibration equations,
the
[(y-y)/y]2
0.00
1.52
1.52
47.09
1.54
1.53
3.92E-06
3.45E-05
97.22
1.56
1.55
148.86
1.59
1.57
9.32E-05
206.58
1.63
1.61
0.000129
270.38
1.67
1.65
0.000218
347.85
1.74
1.71
0.000256
396.46
1.79
1.76
0.000242
455.70
1.85
1.82
0.000225
493.68
1.90
1.87
0.000214
539.25
1.95
1.93
0.000159
listed Table 2.
Data
Usage
575.70
2.00
1.97
0.000151
Head on HS5
0.22 m
Initial condition
622.79
2.06
2.04
7.97E-05
0.30 m
Boundary condition
663.80
2.12
2.10
5.22E-05
0.35 cms*
Initial condition
721.53
2.20
2.19
3.09E-05
RMSE
0.011219(1%)
0.37 m
0.01 cms
Boundary condition
Initial condition
0.58 m
Boundary condition
0.03 cms
Initial condition
0.255 m
Initial condition
0.67 m
Verification
Head on HS1
0.21 m
Verification
6.1
channel
24
6.2
September, 2009
Vol. 2 No.3
Location
Main downstream
0.35
0.35
0.49 m
0.03
0.23 m
0.01
0.37 m
0.4
0.4
1.07 m
0.4
0.2 m
0.22 m
Initial conditioned
0.26 m
0.67 m
0.21 m
The
Comment
Initial condition
0.67 m
The
September, 2009
Vol. 2 No.3
25
The channel
Conclusions
The
1.07 m.
Acknowledgements
The authors thank Mr. Milton Henry, Director of
Engineering & Technical Services, Jamaican National
Irrigation Commission for his work in field survey and
flow measurement.
26
September, 2009
[References]
[1]
[6]
[7]
T.
Open-channel
hydraulics.
New
York:
479.
Chow
[2]
Vol. 2 No.3
[9]
134(1): 3642.
International
Journal
of
Agricultural
&
Dear Dr Yingkuan,
Many thanks for your email, and apologies for the delay
in replying.
Kind regards,
Jason Foulsham
www.theiet.org
United Kingdom
0042-31 14/95/2306-551$6.00
0Swets & Zeitlinger
SUMMARY
This paper provides a broad review of the developments that have taken place within the past thirty
years in the field now known as Advanced Vehicle Control Systems (AVCS). This long time horizon
was chosen to ensure that significant findings from earlier generations of work are not overlooked
today. Following a history of the international development of AVCS, several methods of classifying
AVCS are introduced. The main body of the paper reviews the relevant literature in lateral, longitudinal and integrated control of road vehicles and summarizes the most significant findings from
this work.
The term "Advanced Vehicle Control Systems" (AVCS) has been adopted
within the United States to refer to the subclass of Intelligent Transportation
Systems (ITS) that provide the driver with safety warnings or assistance in
controlling his vehicle, including the extension to full control of vehicle motions.
The term AVCS was defined by the ad-hoc working group that called itself
Mobility 2000, which worked during the period 1988-1991 to establish a U.S.
program for development of ITS. It originally established four functional groups
to define ITS technologies and services (which were known until recently as
Intelligent Vehicle Highway Systems or IVHS):
1.
2.
3.
4.
These functions have since been somewhat redefined and augmented with
additional functional groups. In the original thinking, the distinction between
AVCS and ADIS was based primarily on the applicable time scale, with those
*Deputy Director, California PATH Program, University of California, Berkeley, Institute of
Transportation Studies, Richmond Field Station, 1357 S. 46th Street, Richmond, CA 94804,
U.S.A. Tel: 5 10-23 1-9537, Fax: 510-23 1-9565, e-mail ses@its.berkeIey.edu
552
STEVEN E. SHLADOVER
functions that involve decisions that must be made or actions that must be
implemented in less than 1 second being considered AVCS, while the slower
functions would be ATIS. The "A" in AVCS was later changed to be
"Advanced" in order to be directly comparable to ATMS and ADIS.
The words "Advanced vehicle control systems" are in fact a somewhat
misleading description of the technologies and services that comprise AVCS.
Many of the AVCS systems are not in fact control systems, but rather are systems
that provide warnings of hazardous conditions to drivers or enhance the ability
of drivers to perceive hazards themselves. The emphases in the designs of these
systems are on issues of sensing, sensor data processing and human factors rather
than on vehicle dynamics or control. In light of this, the most recent classification
of ITS user services developed for the U.S. National Program Plan for ITS has
replaced the term AVCS with the more comprehensive term, "Advanced Vehicle
Control and Safety Systems". While AVCS is excessively limiting terminology in
this context, it is also excessively broad in a different sense, because it is not meant
to include a variety of systems that could legitimately be considered to be
advanced vehicle control systems, such as traction control, antilock brakes,
active suspensions and four-wheel active steering systems. These can in some
sense be considered enabling technologies for AVCS but they are not generally
regarded as being AVCS in themselves.
553
and Commerce to study new systems for urban transportation (since the
Department of Transportation did not yet exist). Thls "New Systems Study"
led to the creation of a series of reports from the Stanford Research Institute
(SRI) and General Research Corporation (GRC), culminating in the summary
report, "Tomorrow's Transportation: New Systems for the Urban Future" [6] in
1968. Six new urban transportation systems were recommended for development
in that report, one of which, called "Dual Mode," was essentially the automated
highway using vehicles that could operate in either a manual or an automated
mode (hence the name). The work on dual mode and automated highway
technologies was reported in a Transportation Research Board (TRB) conference in 1974, although the Proceedings [7] unfortunately did not contain full
texts of many of the papers.
In addition to the ongoing work at Ohio State, the FHWA sponsored broader
systems studies of the feasibility of an automated highway system, leading to
reports in 1969 [8] and 1977 [9]which recommended several areas needing further
research. A later FHWA study by General Motors in 1980-82 [lo, 111 examined
the technical potential of the automated highway in the context of realistic
implementation problems. All of these studies addressed broad operating
concepts, system-level tradeoffs and economic issues rather than detailed
engineering designs.
Privately-funded work on the development of roadway automation technologies is not generally reported because of proprietary restrictions on release of
information. This makes it difficult to tell how much activity there actually has
been in this area. Certainly General Motors has been active in the work on
roadway automation technology since the time of their original work, although
only limited glimpses of this have been available to the research community in
general [l2].
A large portion of the technology development work relevant to roadway
automation has actually been performed in pursuit of Automated Guideway
Transit (AGT) systems, formerly referred to as Personal Rapid Transit (PRT)
systems. This activity in the United States was primarily sponsored by the Urban
Mass Transportation Administration (UMTA) during most of the 1970s, and
produced a large body of research results which must be absorbed and understood by anyone hoping to advance the technologies for AVCS. The ACT
researchers had to address most of the issues that must be solved for AVCS,
except for the problems associated with transitions to and from automatic
control and with developing vehicles that can operate under both manual and
automatic control. The results o f t h e AGT research are scattered across an
extensive range of reports, journal papers and conferences. The central references, from which a more comprehensive literature search can be started, are the
four conference proceedings [I 3- 161.
The early developments in the United States were paralleled in Europe and
Japan. In England, the Road Research Laboratory conducted experiments on
automatic steering control in the late 1960s [17], while in Japan MITI's
5 54
STEVEN E. SHLADOVER
555
'
556
STEVEN E. SHLADOVER
services
hierarchical system architecture
a functions
a types of control.
a \ user
The user services breakdown has been used as the organizing principle for the
National Program Plan for Intelligent Transportation Systems in the United
States, under development by ITS America [35]. This breakdown is useful for
communicating how AVCS would be perceived and used by the general public,
rather than for addressing in-depth technical issues. The current listing of
relevant user services under consideration in the U.S. is:
Advanced Vehicle Control and Safety Systems ( A VCSS)
Longitudinal Collision Avoidance
Rear-End Crash Warning and Control
Autonomous Intelligent Cruise Control (AICC)
Cooperative Intelligent Cruise Control (CICC)
Head-On Crash Warning and Control
Passing Warning and Control
Backing Crash Warning
558
STEVEN E. SHLADOVER
A second way of subdividing AVCS is to consider the information architecture that could be used to implement the AVCS services. This would be
transparent to the transportation system users, but is critically important to
system developers. Varaiya has proposed a hierarchical framework for organizing the design of an automated highway system, which has also proven to be
extremely useful for structuring thought about all ITS system designs [36]. This
framework disaggregates system elements in terms of their information span,
frequency of update of data, spatial scope and function. In this scheme,
information is exchanged between adjacent layers or among peer entities at
the same layer, providing a way of simplifying system design and clarifying the
functions that must be implemented by each entity. The five layers of the
architecture are:
1. Network - entire urban or regional transportation network
2. Link - section of road of the order of 1 km in length
3. Coordination - cluster of neighboring vehicles that must coordinate
motions to ensure safety
4. Regulation - closed-loop control implementation on an individual vehicle
5. Physical - sensors and actuators on individual vehicle.
The issues of vehicle dynamics and control are normally confined to the
regulation and physical layers within this architecture. The large majority of
the existing literature addresses these layers, since the higher layers are only
relevant to the most sophisticated and long-term system designs. However, the
literature on the higher layers is starting to grow vigorously.
Within the regulation and physical layers, the functions that must be
performed by an AVCS system can be subdivided using traditional block
diagram schematics of feedback control systems. The fully automated systems
are somewhat different from the warning and assistance systems, but the overall
schematics remain quite similar. Figure 1 illustrates the general functional block
diagram for AVCS warning and assistance systems, while Figure 2 shows the
analogous diagram for fully-automated AVCS systems. Each diagram includes
Raw Sensor
Sensor
Dynamics
Driver Actions
(Human Factors)
Driver
Interface
Fig. I(a). Functional block diagram for driver warning and assistance systems.
Sensors
Sensor
homing
Corn utotion
of L z a r ~
Warning
Driver
Interface
Human
Factors
Vehicle
Dynamics
Fig. I(b). Relative levels of technical challenges for driver warning and assistance systems.
Desired
Path
Sensor
Control
Control
Command
Actual
Trajectory
Actuation
Fig. 2(a). Functional block diagram for automated systems.
560
STEVEN E. SHLADOVER
Senran
Sensor
Processing
Control
Computation
Vehicle
Dynamics
Control
Actuation
56 1
562
STEVEN E. SHLADOVER
based on the vehicle's own power steering system [17] and a Ford Cortina using
an electromechanical servo system [42]. Both steering servo actuation systems
proved to be problematic in this research. RRL made a serious commitment to
automatic lateral control research, as evidenced by their installation of 16 km of
reference cable in the outer lane of the M4 motorway near Reading, England [43].
563
processing software used to identify the location and alignment of the vehicle
relative to the lane markers [49,50]. This work was derived directly from
technology developed for factory automation, where speeds are low and vehicle
dynamic effects are not strong.
All of the approaches mentioned thus far were based on use of passive sensors
mounted on the automatically steered vehicles. An additional concept, developed by Spangler and Snell[51] and elucidated by TRW [8], was based on the use
of a radio frequency transmitter on the vehicle inducing currents in a passive
conductor beneath the road surface, and inductive coils on the vehicle then
sensing the magnetic field produced by these currents. This system was of course
very sensitive to cross-talk between the transmitter and the sensor coils.
A variety of automatically steered vehicle applications were reported in the
1960s and early 1970s, but none of them seemed to endure for long. These
included an electronic steering control system developed for a full-size passenger
bus by Barrett Electronics in 1961 [52], and vehicles that operated at very low
speeds in hospitals and warehouses [53]. The sensing aspect of lateral guidance
was used to provide position error information to human drivers steering
military vehicles through Arctic wastelands in adverse weather [54].
Much research was done in the 1960s and 1970s on mechanical guidance of
automatically steered vehicles, for application to automated guideway systems
and busways. Various of these systems have been implemented and are operating
in public service today, with performance that is in some cases excellent. This
work is not addressed here because its dependence on the availability of a special
mechanical reference is not compatible with AVCS needs to operate in more
general roadway conditions.
564
STEVEN E. SHLADOVER
The M.I.T. lateral control research continued with use of optimal linear
regulator theory for lane following [59], and also addressed transient effects
associated with curve entry [60]. A very similar technical approach was adopted
by Panik and his colleagues at Daimler Benz [61], with the important addition of
an experimental verification of the control system design on a full-scale
passenger bus, using an electronic wire guidance reference [62,63]. This work
is one of the most important contributions to lateral control because it has been
extended to operational tests in actual passenger service in Fiirth, Germany
(electronic reference), as well as being used in daily transit operations in Essen,
Germany and Adelaide, Australia (mechanical reference). The electronic
reference system has more recently been applied for lateral control of the
special-purpose vehicles that operate in the service tunnel of the Eurotunnel
(the English Channel tunnel).
565
which were evaluated using simulations before being tested on vehicles. The
analytical basis for this work, including the feedforward of roadway curvature
information, is described by Peng and Tomizuka [71,72]. They used a frequencyshaped linear quadratic (FSLQ) optimal controller to explicitly address the
maximization of ride quality, while the roadway geometry preview helped keep
the feedback gains low. Peng et al. [73,74] conducted experiments in cooperation
with IMRA America to demonstrate the performance of this controller and to
compare it with a simpler PID with preview controller. These experiments
showed excellent ride quality and curve tracking accuracy, including fairly
aggressive curving (0.27 g lateral acceleration) and anomalous operating conditions (extra payload weight, partially deflated tires in different positions, snowy
road surface). The lateral tracking errors remained less than 10 cm under almost
all conditions, and less than 15cm in the worst cases. Very significantly, the
performance was shown to be measurably better than the steering of the same
test vehicle by a variety of human drivers [73], and the control system initially
designed for a Toyota Celica was converted for use on a Pontiac 6000 with a
minimum of modifications, while retaining excellent performance [74].
The PATH lateral control work has been extended to testing of an alternative
controller, using fuzzy control [75]. Controller design has been initiated for lane
changing maneuvers [76], using modified FSLQ and sliding controllers, but this
has not yet been proven experimentally. Two other lateral control design efforts
within PATH have been based on the idea of vehicles sensing the movements of
the vehicles in front of them, rather than sensing their location relative to the
roadway lane center (in order to save the cost of any special roadway
infrastructure). Narendren and Hedrick proposed "auto-tracking" using
vehicle-to vehicle range and azimuth angle measurements and a sliding control
law [77]. Fujioka proposed a trajectory-following controller, also using sliding
control, in which the preceding vehicle would need to communicate its tracking
error to the following vehicle [78]. Neither of these approaches is applicable to
the first vehicle in sequence (which would have no other vehicle to follow) and
neither has yet been verified experimentally.
Several researchers have addressed the robustness of lateral control systems
from different perspectives. Ackermann (79-811 has emphasized analytical
approaches to deriving fixed-gain controllers that can steer successfully under
a full range of speed, payload and road surface conditions and has shown
analysis and simulation results with good performance in the absence of preview
information about road geometry [80,81]. Patwardhan and Tomizuka [82] have
designed a control system that can be reconfigured to accommodate the change
in vehicle dynamics produced by a sudden tire blow-out. They have tested their
open-loop vehicle model against experimental results using a full-scale vehicle
with a sudden tire deflation, but have not yet tested the closed-loop controller.
Smith and Starkey [83] studied automatic lateral control for emergency
maneuvers that would require higher lateral accelerations than contemplated
in the designs of other lateral control systems. They showed howgain scheduling
566
STEVEN E. SHLADOVER
567
markers. Yasui and Margolis [93] described the design of a system that would use
retroflective markers between the lanes as the reference for an optical scanning
sensor, providing both preview and current vehicle orientation information to a
simulated lateral control system. Zhang and Parsons [94] proposed the use of
discrete permanent magnets as the reference markers for lateral control, and
conducted a series of experiments to prove that these could provide extremely
accurate (5 mm) reference position information to a vehicle equipped with
magnetometer sensors. These magnetic markers, which can be installed in a
sequence that incorporates a binary code (N or S poles up) to represent
upcoming road geometry changes, have served as the reference for most of the
PATH lateral control experimental research.
5.9. Summary of Lateral Control State of Knowledge
Based on the research conducted thus far, the current state of knowledge about
automatic lateral control appears to include the following:
568
STEVEN E. SHLADOVER
569
570
STEVEN E. SHLADOVER
57 1
572
STEVEN E. SHLADOVER
CONTROL STRUCTURE NUMBER
ROADSIDE
ommanded speed
LEADING VEHICLE
PRECEDING VEHICLE m
pacing to preceding vehicle m + 1
PRECEDING VEHICLE 1
acing to preceding vehicle 2
ommanded speed
TRAILING VEHICLE
pacing to controlled vehicle
peed
cceleration
TRAILING VEHICLE n
Fig. 3. Sources of feedback information for vehicle-follower longitudinal control (variables available by autonomous sensing listed by bold face).
linear, time-invariant control system design approaches. By the late 1970s it had
become apparent that these were not adequate for addressing the important
performance issues, and from that time forward most relevant work has included
nonlinear, time varying and adaptive control considerations. Very few of these
structures have been implemented and verified experimentally on full-scale
vehicles (structures 2,3,4,8,9).
One of the most persistent and controversial issues in vehicle-following
control systems is the selection of the desired following distance. The three
basic classes of vehicle following are:
574
STEVEN E. SHLADOVER
575
human interface issues rather than on vehicle dynamics or control law design.
Technical issues were addressed more specifically by Sala and Pressi [130], who
addressed the different spacing criteria that were considered for use in the
PROMETHEUS AACC work. This included the general categories that were
already mentioned here in Section 6.3, and provided more specifics on how the
safe following distances were to be estimated. This paper also defined strategies
for cooperative ACC, in which the vehicles are linked by communication systems
as well.
Work on AACC in Japan and the U.S. has lagged somewhat behind that in
Europe. Satoh and Taniguchi [131] described an AACC based on use of a laser
radar sensor and a controller designed to try to approximate normal driver
behavior. Their experiments on the test vehicle showed that drivers were quite
comfortable with the system's responses, but the paper gave no indication of
quantitative effects on traffic flow or the stability of this type of controller when
operated on multiple consecutive vehicles. Ioannou and his students have
applied this AACC control structure to the design of controllers for sequences
of fully automated vehicles operating in reserved lanes [132,133]. They have
shown that they can achieve asymptotic string stability with this simple feedback
structure by applying the constant (time) headway vehicle following criterion.
Their extensive controller design and simulation work has been followed with
some experiments implementing the controllers on test vehicles [134].
The largest body of technical papers on AACC has been produced by Paul
Fancher and his colleagues at the University of Michigan, working in cooperation with Leica, who developed the prototype hardware with which they have
been experimenting [135- 1381. They described their controller as fundamentally
a PI compensator with a limit on integrator wind-up [135]. Their infrared sensor
measures the range to the preceding vehicle, and they derive range rate from that
range information. They very carefully considered how to design the controller
for operation throughout the phase plane defined by range and range rate
measurements [136]. They evaluated the likely effect of the AACC on traffic flow
by means of simulations of deceleration maneuvers by strings of AACC vehicles.
The controller design focused on choosing the most appropriate vehiclefollowing time gaps (loosely speaking, headways) based on the control system
time constants, and concluded that the AACC would show asymptotic string
stability if the time gap is less than twice the control system time constant (lag)
for a constant-headway vehicle following mode of operation [137]. Experiments
with a variety of drivers in the equipped test vehicle highlighted the need to be
able to adjust performance.to reflect differing driver preferences [138].
576
STEVEN E. SHLADOVER
577
without the acceleration feedback. The controller with this higher-order feedback also generates a higher-order control command (a jerk command as
compared to an acceleration command for the other controller). This controller
was shown to be vulnerable to instability because of the jerk-limiting saturation
nonlinearity.
6.3.6. Feedback of the "Complete" State of the Preceding Vehicle and Controlled
Vehicle Plus Commanded Velocity of Preceding Vehicle (Structure 6 )
This structure was proposed by Tsugawa [146,147]as part of his concept of "soft
linked" vehicles. This concept was proposed for implementation using communications between vehicles rather than range or range rate sensing. Each vehicle
would need to know its own absolute position in order to avoid the need to
measure the spacing between them. This was originally proposed for very lowspeed use within factories or warehouses, with infrared communications between
successive vehicles, but then was extended to road vehicle applications. It was
simulated, but does not appear to have been implemented experimentally on
road vehicles.
6.3.7. Feedback of the "Complete" State of the Preceding Vehicle and Controlled
Vehicle Plus Jerk of Preceding Vehicle (Structure 7 )
This structure was analyzed in depth by Pue of the Applied Physics Laboratory
in the late 1970s [148- 1511as a way of overcoming the awkwardness of the timevarying gains in the work by Chiu, Stupp and Brown. Pue developed a family of
rather complicated nonlinear controllers that explicitly incorporated the kinematic boundary of his colleagues and state constraints as well. His optimum and
one of his suboptimum controllers used this information structure, while he also
studied two other suboptimum controllers of structure 5. By means of simulations and describing function analysis he was able to demonstrate asymptotic
stability and good performance of this controller on a triple integration vehicle
model, with no propulsion system dynamics. Since this approach requires
communication of data that cannot be measured directly, these studies included
consideration of the needed data rates and the effects on performance of
communication imperfections (sampling, quantization and delays) [15 11.
'
578
STEVEN E. SHLADOVER
6.3.9. Feedback of the "Complete" State of the Controlled Vehicle and its
Predecessor, Plus the Velocity andAcceleration of the Platoon Leader (Structure 9 )
This extension of feedback structure 8 was developed by researchers working in
the PATH Program, to provide more flexibility in defining the closed-loop
dynamics of the complete multi-vehicle coupled platoon with constant intraplatoon spacings. The addition of acceleration feedback makes it possible to
place the zeros as well as the poles of the closed-loop system transfer functions,
although it does require additional communication of data among the vehicles.
This has been applied to relatively simple third-order vehicle dynamic models
that incorporate some limited nonlinear effects [153,154] as well as models that
include explicit representations of powertrain dynamics [155-1571. The controller design approaches have included linear pole-placement [153,154] and
nonlinear sliding control [155- 1571.
These analysis and simulation studies demonstrated asymptotic stability of
the platoon of vehicles under relatively idealized conditions, as well as in the
presence of real-world imperfections such as process and measurement noise,
and quantization and sampling errors [153]. A companion study [I581 illustrated
the moderate loss of asymptotic stability that resulted from the loss of the
communication of lead vehicle information, again under idealized conditions,
while another [I591 showed the effects of communication delays, differences in
vehicle performance, and measurement noise.
This family of controller, building on the sliding control designs of Hedrick
and his colleagues described in 1155-1571, has been evaluated in extensive
experiments by PATH Program researchers, beginning with two vehicles and
extending to the control of four-vehicle platoons. Initial experiments were not
able to conclusively show asymptotic stability because of ranging sensor
limitations [160], but more recent experiments with an improved ranging sensor
have produced superior results, with excellent tracking accuracy and ride quality
[161,162]. The earlier experiments in this series illustrated the importance of
having fast throttle actuator response in order to achieve stability of vehicle
following (even with only two vehicles). The latest experiments have demonstrated the ability to operate four vehicles at a constant spacing of 4 m with
variations within 30cm, while executing a driving cycle including moderate
accelerations (0.1 g) and decelerations (0.05 g) on level roadway as well as on
grades of up to 3%. These experiments have been conducted using throttle
control but no brake control, because related studies of braking actuation have
shown the inadequate dynamic response of the brake actuator system that acts
on the brake pedal (thereby including brake booster dynamics) [163].
6.3.10. Feedback of the "Complete" State of the Controlled Vehicle and Its N
Predecessors (Structure 10)
This structure has been termed "Continuous Platooning" by Ren and Green
[164], whose analyses have indicated its advantages in responding to disturbances several vehicles ahead of the controlled vehicle (but not as far ahead as a
579
580
STEVEN E. SHLADOVER
58 1
and control technology, interest has focused much more heavily on vehiclefollower longitudinal control systems than on point-follower or fixed block
systems in recent years. Although these systems require more complicated
technology to be implemented on the vehicles than the others, their
advantages in terms of overall system capacity (vehicles per hour), fault
tolerance, and total system cost outweigh that concern.
a The longitudinal dynamics of light-duty vehicles powered by spark ignition
engines with electronic fuel injection do not appear to be limitations to the
achievement of high performance longitudinal control systems, with
smooth and precise control of spacings between vehicles.
a A wide variety of possible feedback control structures have been proposed
and evaluated over the years, although only a few of them have been
verified experimentally. A few of these structures can be implemented on
autonomous vehicles (using vehicle-mounted sensors only, without communication), but most of them require communication of state information
from other vehicles.
a There are limits to the performance that can be achieved at each level of
feedback information (or control structure), and higher levels of performance generally require more information.
a Vehicle-follower systems should be designed to have asymptotic string
stability for an arbitrary length string of vehicles. Designs that are stable for
one vehicle following another are not necessarily asymptotically stable for
larger numbers of consecutive vehicles. Asymptotic stability needs to be
assured under all realistic abnormal operating conditions, including
communication delays, sensor noise, sampling and quantization errors,
parameter variations, etc. ,
a Different criteria can be applied to determine the relationship between
582
STEVEN E. SHLADOVER
583
584
STEVEN E. SHLADOVER
also reported as early as 1990 [192], with an emphasis on design to optimize ride
comfort for the vehicle passengers in automatic lane change maneuvers.
Toyota's research, on fully automated driving concentrated initially on testing
of unmanned vehicles on the company's test track, rather than on public roads
[193]. More recently, they have extended their work to the development of
systems that could be applied to automate the driving of customers' vehicles on
suitably equipped roads (using inductive cable guidance combined with stereo
vision for lane detection and a laser radar for obstacle detection/vehcle
following) [194]. Probably the most sophisticated such development to date
has been the work of Daimler Benz on the VITA I1 test vehcle, an automobile
equipped with 12 video cameras and special-purpose image processing computers to handle all aspects of regulation-layer driving [195]. This vehicle incorporates lane tracking, vehicle following and forward and side obstacle detection
and avoidance, and has been driven under automatic control on public highways
in Germany.
585
Above the coordination layer is the link layer, where the relevant variables are
associated with the flows of vehicles rather than the motions of individual
vehicles. The earliest treatment of the issues that arise at this layer appears to be
by Karaaslan et a1 [205], who addressed the capacity of an automated lane and
the ways in which wayside recommendations of speed could be used to attenuate
flow disturbances. These issues were explored in much greater depth by Rao and
Varaiya, who developed heuristics for assigning vehicles to lanes within an
automated roadway facility under both normal and incident conditions
[206,207]. These papers showed very good system performance with the
heuristic control approach. Ramaswamy et a1 [208] have taken .a different
approach to automatic lane assignment, based on a centralized optimization
of trajectories that appears to be very communication intensive. Peng has
adopted a more traditional optimal control approach to the link-layer traffic
management problem [209].
The highest layer of the fully automated control system is the network layer.
Some of the issues at this layer were addressed in studies of automated guideway
transit systems in the 1970s, and Fenton's group at OSU addressed them for
automated highways at that time as well [4]. Those earlier studies must now be
considered obsolete because of the dramatic changes in computer hardware and
software and communications technologies that have occurred since then. The
first tentative modern approach to the issues encountered at this layer appears to
have been by Tsao [210], but much more remains to be done here.
9. CONCLUDING REMARKS
AVCS is an example of a field of activity in which we need to repeatedly remind
ourselves that there are few completely new ideas. There is a heritage of more than
thirty years of applicable research upon which to build. During this time, much
has been learned about concepts and technical approaches, both favorable and
unfavorable. That body of knowledge has been documented to a considerable
extent, and is available to researchers who are interested in working on AVCS.
This paper has attempted to summarize the highlights of what has been
accomplished in AVCS. It is not meant to be an exhaustive survey of AVCS, and
indeed it does not even address many aspects of AVCS. However, it should
provide reasonably complete coverage of the state of knowledge about the
vehicle dynamics and control aspects of AVCS. Readers who take the time to
consult the references cited herein should find that a worthwhile investment, one
that will save them much time and wasted effort.
ACKNOWLEDGMENT
This work was performed as part of the California PATH Program of the
586
STEVEN E. SHLADOVER
REFERENCES
I . Gardels, K.: Automatic car controls for electronic highways. General Motors Research
Laboratories Report GMR-276, 1960.
2. Zworykin, V.K., and Flory, L.E.: Electronic control of motor vehicles on the highways. Highway Research Board Proceedings, 37th Annual Meeting, 1958.
3. Barrick, D.: Automatic Steering Techniques. IRE Inter. Conv. Rec. 10, Part 2 (1962), pp. 166178.
4. Fenton, R.E., and Mayhan, R.J.: Automated highway studies at the Ohio State University An overview. IEEE Trans. on Veh. Tech. 40,l (1991), pp. 100-1 13.
5. Hanson, M.E. (Ed.): Project METRAN: An Integrated, Evolutionary Transportation System
for Urban Areas, MIT Press, 1966.
6. Tomorrow's transportation: new systems for the urban future. U.S. Dept. of Housing and
Urban Development, 1968.
7. Dual mode transportation. Proc., conference conducted by Transportation Research Board,
May 29-31, 1974 TRB Special Report 170, 1976.
8. TRW Systems Group: Automated Highway Systems, December 1969, 572 pp. (PB 191 696).
9. Calspan Corp.: Practicability of automated highway systems. Report DOT-FH-11-8903,
1977.
.. . ..
10. Bender, J.G., et al.: System studies of automated highway systems. General Motors Transportation Systems Center Report EP-81041A. 1981; FHWA Report No. FHWAIRD-821
003, 198i.
11. Bender. J.G.: An overview of svstems studies of automated hiehwav svstems. IEEE Trans.
~ e hTech.
.
40,l (1991), p i . 82-99.
12. Smisek, R.R., and Harder, G.A.: Automated control of guideway transit vehicles. 29th IEEE
Vehicular Technology Conference, Arlington Heights, IL, March 1979, pp. 57-64.
13. Anderson, J.E., et al. (Eds.): Personal rapid transit. Institute ofTechnology, University of Minnesota, 1972.
14. Anderson, J.E. (Ed.): Personal rapid transit 11. University of Minnesota, 1974.
15. Gary, D.A. (Ed.): Personal rapid transit 111. University of Minnesota, 1976.
16. Advanced Transit Association: International Conference on Advanced Transportation and
Urban Revitalization, Indianapolis, IN, May 1978.
17. Cardew, K.H.F.: The Automatic Steering of Vehicles - An Experimental System Fitted to a
DS Citroen Car, Road Research Laboratory Report RRL LR 340, Crowthorne, England, 1970, 29 pp. (PB 196 983).
18. Oshima, R., et al.: Control system for automobile driving. Proc. Tokyo IFAC Symposium
(1965), pp. 347-357.
19. Fujioka, T., Yoshimoto, K., and Takaba, S.: A case study on an automated driving highway
system in Japan. Transportation Research Board 72nd Annual Meeting, Washington,
D.C., January 1993.
20. Hesse, R.: Cabinentaxi - A personal public transportation system. Transp. 1,3 (1972), pp. 321329.
21. Ishii, T., et al.: The control system of CVS using the two-target tracking scheme. personal
Rapid Transit 11, Dept. of Audio Visual Extension, University of Minnesota (1974),
pp. 325-334.
- - -
587
22. Maury, J.P.: The ARMIS PRT system. SAE Paper 740143, 1974.
23. Broqua, F., Lerner, G., Mauro, V., and Morello, E.: Cooperative driving: basic concepts and a
first assessment of "intelligent cruise control" strategies. Proc. Advanced Telematics in
Road Transport, Vol. 11, DRIVE Conference, Brussels, February 1991, pp. 908-929.
24. de Saint Blancard, M., Dubuisson, B., and Rombaut, M.: PRO-LAB2: A vehicle with driving
assistance. Proc. First World Congress on Applications of Transport Telematics and
Intelligent Vehicle-Highway Systems, Paris, November-December 1994. pp. 1871 - 1878.
25. Saxton, L.: Mobility 2000 and the roots of IVHS. IVHS Rev. (Spring 1993), pp. 11-26.
26. Shladover, S.E., Bushey, R., and Parsons, R.E.: California and the roots of IVHS. IVHS Rev.
(Spring 1993), pp. 27-34.
28. Tsugawa, S., Watanabe, N., and Fujii, H.: Super smart vehicle system - its concept and preliminary works. Proc. Vehicle Navigation and Information Systems Symposium, Dearborn, MI, October 1991, pp. 269-277 (SAE Publication P-253).
29. Nakamura, M., Kawashima, H., Yoshikai, T., and Aoki, K.: Road-vehiclecooperation driving
system. Proc. 1994 Vehicle Navigation and Information Systems, Yokohama. Japan,
August-September 1994, pp. 425-430.
30. Leasure, W.A., and Burgett, A.L.: NHTSA's IVHS collision avoidance research program: strategic plan and status update. Proc. First World Congress on Applications of Transport
Telematics and Intelligent
Paris, November-December 1994,
- Vehicle-Highway
- Systems,
pp. 22 16-2223.
31. Shladover. S.E.: The California PATH Droaram of IVHS research and its aporoach to vehiclehighhay automation. Proc. lnteliigent Vehicles 92 Symposium, ~ e t i d i t MI,
, June-July
1992, pp. 347-352.
32. Chang, K.S., Hedrick, J.K., Zhang, W.-B., Varaiya, P., Tomizuka, M., and Shladover, S.E.:
Automated highway system experiments in the PATH program. IVHS J. 1,1 (1993),
pp. 63-87.
33. Saxton, L., and Bishop, J.R.: The United States DOT automated highway system (AHS) program. Proc. First World Congress on Applications of Transport Telematics and Intelligent Vehicle-Highway Systems, Paris, November-December 1994, pp. 1952-1958.
34. Bishop, J.R., and Stevens, W.: Results of precursor systems analyses of automated highway
systems. Proc. First World Congress on Applications of Transport Telematics and Intelligent Vehicle-Highway Systems, Paris, November-December 1994, pp. 1959-1966.
35. National Program Plan for Intelligent Transportation Systems, Final Draft, ITS America,
November 1994.
36. Varaiya, P., and Shladover, S.E.: Sketch of an IVHS Systems Architecture, Proc. Vehicle Navigation and Information Systems Conference, Paper No. 912838, Dearborn, MI, October
1991, pp. 909-928.
37. Zhang, W.-B., Shladover, S.E., Hall, R.W., and Plocher, T.: Functional definition of automated highway systems, TRB Paper 940988, 73rd Annual Meeting, January 1994,
Washington, D.C.
38. Transportation Research Institute, Carnegie-Mellon University, Urban Rapid Transit Concepts and Evaluation, TRI Research Report #I, Pittsburgh, 1968, 241 pp.
39. Bidwell, J.B., and Cataldo, R.S.: Electronic chauffeur may guide car of future. SAE J., (September 1958), pp. 64-67.
40. Electronic control of motor vehicles on the highways. Proc. Highway Research Board 37th
Annual Meeting, 1958, pp. 436-451.
41. Flory, L.E., Gray, Morey, Pike, and Caulton: Electronic techniques in a system of highway
vehicle control. RCA Rev. 23,3 (1962), pp. 293-310.
42. Penoyre, S.: Dual mode road vehicles. Traff. Eng. Contr. 13,s (1971), London, pp. 189-191.
43. Road Research 1971. Annual Report of the Road Research Laboratory, London: Her Majesty's Stationery Office, 1972, 120 pp.
44. Fenton, R.E., Olson, K.W., and Bender, J.G.: Advances Toward the Automatic Highway.
Highway Research Record #344, 1971, pp. 1-20 (including commentary by Wilkie,
Lenard, and Segal).
45. Olson, K.W., Sapp, and Flaig: A System for Automatic Vehicle Lateral Guidance, Highway
Research Record #275, 1969, pp. 1-1 1.
46. Olson, K.W.: Wire-reference configurations in vehicle lateral control. IEEE Trans Veh. Tech.
VT-26,2 (1977), pp. 161-172.
588
STEVEN E. SHLADOVER
47. Fenton, R.E., Melocik, G.C., and Olson, K.W.: On the steering of automated vehicles: Theory
and experiment. IEEE Trans. Auto. Control AC-21,3 (1976), pp. 306-3 15.
48. Fenton, R.E., and Selim, 1.: On the optimal design of an automotive lateral controller. IEEE
Trans. Veh. Tech. VT-37,2 (1988), pp. 108-1 13.
49. Tsugawa, S., Yatabe, T., Hirose, T., and Matsumoto, S.: An automobile with artificial intelligence. Proc. Sixth International Joint Conference on Artificial Intelligence, Vol. 2,
Tokyo, Japan, August 1979, pp. 893-895.
50. Tsugawa, S.: Vision-based vehicles in Japan: the machine vision systems and driving control
systems. Proc. IEEE International Symposium on Industrial Electronics, Budapest,
Hungary, June 1993.
51. Spangler, R.A., and Snell, F.M.: A New Vehicle Guidance and Speed Control System, Highway Research Bulletin #338, September 1962, pp. 20-27.
52. Driverless Bus Tested. Automotive Industries, 125,9 (1960, p. 29.
53. Richards, 6.: New Movement in Cities, London: Studio Vista, 1966, 96 pp.
54. O'Rourke, C.O.: Electronic trail-finding. Contr. Eng. 4,5 (1957), pp. 117-1 19.
55. Pasternack, S.: Modern Control Aspects of Automatically Steered Vehicles, DOT-TSC-OST72-3, Technical Note, December 1971, 16 pp.
56. Pasternack, S.: Some modern control aspects of automatically steered rubber tired vehicles,
ASME Paper No. 73-ICT-54, Second lntersociety Conference on Transportation, Denver, September 1973.
57. Bonderson, L.S.: Optimal lateral control for dual-mode vehicles. Dual-Mode Conference,
Washington, D.C., May 1974.
58. Shladover, S.E.: Performance limits of automatic steering controllers. Gary, D.A. (Ed.): Personal Rapid Transit 111. University of Minnesota, Minneapolis, M N , 1976, pp. 233-245.
59. Shladover, S.E., Wormley, D.N., Richardson, H.H., and Fish, R.: Steeringcontroller design for
automated guideway transit vehicles, J. Dyn. Sys. Meas. Contr. 100,l (1978), pp. 1-8.
60. Shladover, S.E., Fish, R., Wormley, D.N., and Richardson, H.H.: Performance of steering
controllers for automated guideway transit vehicles. Proc. 5th Vehicle System Dynamics
- 2nd International Union of Theoretical and Applied Mechanics Symposium:
Dynamics of Vehicles on Roads and Tracks, Vienna, September 1977. Swets and Zeitlinger B.V., Amsterdam, pp. 127-135.
61. Christ, H., Darenberg, W., Panik, F., and Weidemann, W.: Automatic track control of
vehicles. Theory and experiment. Slibar, A,, and Springer, H. (Ed.): The Dynamics of
Vehicles on Roads and Tracks. Swets & Zeitlinger, Amsterdam, 1978, pp. 145-164.
62. Darenberg, W.: Automatische spurfiihrung von Kraftfahrzeugen. Automobil-Industrie (February 1987).
63. Weidemann, W.: Electronic lateral control of vehicles. SAE Technical Paper Series, No.
881 166, SAE, Warrendale, PA, Future Transportation Technology Conference and
Exposition, San Francisco, CA, August 1988.
Computer vision applied to vehicle operation. SAE Technical Paper Series,
64. Metzler, H.-G.:
No. 881 167, SAE, Warrendale, PA, Future Transportation Technology Conference
and Exposition, San Francisco, CA, August 1988.
65. Dickmanns, E.D., and Zapp, A.: A curvature-based scheme for improving road vehicle guidance by computer vision. Mobile Robots. Proc. SPIE, Vol. 727 (1986), pp. 161-168.
66. Dickmanns, E.D., and Zapp, A,: Autonomous high-speed road vehicle guidance by computer
vision. Tenth IFAC World Congress, Munich, 4 (1987), pp. 232-237.
67. Graefe, V.: Vision for intelligent road vehicles. Proc. Intelligent Vehicles '93 IEEE Symposium,
Tokyo, July 1993, pp. 135-140.
68. Masaki, I. (Ed.): Vision-Based Vehicle Guidance, Springer Verlag, New York, 1992.
, ~ u n e1992.
69. Proc. intelligent Vehicles '92 Symposium, IEEE, ~ e t r o sMI,
70. Proc. lntellieent Vehicles '93 Svmoosium. IEEE. Tokvo. Julv 1993.
71. Peng, H., an; Tomizuka, M. vehicle lateral control for-highway automation. Proc. 1990 American Control Conference, San Diego, May 1990, pp. 788-794.
72. Peng, H., and Tomizuka, M.: Preview control for vehicle lateral guidance in highway automation. Proc. 1991 American Control Conference, Boston, MA, June 1991, pp. 3090-3095.
73. Peng, H., Zhang, W.-B., Shladover, S.E., Tomizuka, M., and Arai, A,: Magnetic-marker-based
lane keeping: a robustness experimental study. SAE Congress, Detroit, MI, February
1993, SAE Paper No. 930556, pp. 127-132.
589
74. Peng, H., Zhang, W.-B., Tomizuka, M., and Shladover, S.E.: A reusability study of vehicle
lateral control system. Veh. Sys. Dyn. 23,4 (1994), pp. 259-278.
75. Hessburg, T., and Tomizuka, M.: Fuzzy logic control for lateral vehicle guidance. IEEE Contr.
Sys. Mag. 14,4 (1994), pp. 55-63.
76. Chee, W., and Tomizuka, M.: Lane change maneuver for AHS applications. Proc. International Symposium on Advanced Vehicle Control 1994, Japan, October, 1994, p. 420425.
77. Narendran, V.K., and Hedrick, J.K.: Autonomous lateral control of vehicles in an automated
highway system. Veh. Sys. Dyn. 23,4 (1994), pp. 307-324.
78. Fujioka, T., and Suzuka, K.: Lateral autonomous driving by sliding control. Proc. International Symposium on Advanced Vehicle Control 1994, Japan, October, 1994, pp. 432437.
79. Ackermann, J., and Sienel, W.: Robust control for automatic steering. Proc. 1990 American
Control Conference, San Diego, May 1990, pp. 795-800.
80. Guldner, J., Utkin, V.I., and Ackermann, J.: A sliding mode control approach to automatic car
steering. Proc. American Control Conference, Baltimore, MD, June 1994, pp. 19691973.
81. Ackermann, J., Guldner, J., Sienel, W., Steinhauser, R., and Utkin, V.I.: Linear and nonlinear
controller design for robust automatic steering. IEEE Trans. on Contr. Sys. Tech. (1994).
82. Patwardhan, S., Tomizuka, M., Zhang, W.-B., and Devlin, P.: Theory and experiments of tire
blow-out effects and hazard reduction control for automated vehicle lateral control
system. Proc. International Symposium on Advanced Vehicle Control 1994, Japan,
October, 1994, pp. 426-43 1.
83. Smith, D.E., and Starkey, J.M.: Effects of model complexity on the performance of automated
vehicle steering controllers: controller development and evaluation. Veh. Sys. Dyn. 23,8
(1994), pp. 627-645.
84. Kushiro, I., Kawakami, S., and Saitou, R.: Crosswind feedforward control by active front
steering. Proc. International Symposium on Advanced Vehicle Control, Yokohama,
Japan, September 1992, pp. 402-406.
85. Adachi, T., Butsuen, T., and Kamiyama, S.: Application of real-time simulation to the lanekeeping system. Proc. 1994 Vehicle Navigation and Information Systems, Yokohama,
Japan, August-September, 1994, pp. 415-418.
86. Lemarchand, C., and Coffin, F.: Conception method of man machine interface for an automotive vehicle transversal driving assistance system. Proc. First World Congress on Applications of Transport Telematics and Intelligent Vehicle-Highway Systems, Paris,
November-December 1994, pp. 1820-1827.
87. Mayhan, R.J., and Bishell, R.A.: A two-frequency radar for vehicle automatic lateral control.
IEEE Trans. Veh. Tech. VT-3 1,l (1982), pp. 32-39.
88. Giubbolini, L., and Cugiani, C.: Radar systems for the control of the lateral position of the
vehicle traveling along a motonvay. Proc. International Symposium on Automotive
Technology and Automation, Aachen, Germany, 1993, pp. 689-696.
89. Cugiani, C., and Giubbolini, L.: Millimetre wave radar sensor for the highway global positioning of a vehicle. Proc. 1994 Vehicle Navigation and Information Systems, Yokohama,
Japan, August-September, 1994, pp. 409-414.
90. Johnson, B., et al.: A lateral guidance radar module. Proc. International Symposium on Automotive Technology and Automation, Aachen, Germany, 1993, pp. 697-703.
91. Bishel, R.A.: Dual-mode truck: automated and manual operation. SAE Future Transportation
Technology Conference, 1994.
92. Johnston, A.R., Assefi, T., and Lai, J.Y.: Automated vehicle guidance using discrete reference
markers. IEEE Trans. Veh. Tech. VT-28,l (1979), pp. 95-106.
93. Yasui, Y., and Margolis, D.L.: Lateral control of automobiles using a look-ahead sensor. Proc.
International Symposium on Advanced Vehicle Control, Yokohama, Japan, September
1992, pp. 292-297.
94. Zhang, W.-B., and Parsons, R.E.: An intelligent roadway reference system for vehicle lateral
guidance/control. Proc. 1990 American Control Conference, San Diego, May 1990,
pp. 28 1-286.
95. Chu, K.-C.: Decentralized control of high-speed vehicular strings. Trans. Sci. 8 (1974), pp.
361-384.
590
STEVEN E. SHLADOVER
96. Rahimi, A., Hajdu, L.P., and Macomber. H.L.: Safety analysis for automated transportation
systems. Trans. Sci. 7,l (1973). pp. 1-17.
97. Stepner, D.E., Hajdu, L.P., and Rahimi, A.: Safety considerations for personal rapid transit.
Personal Rapid Transit, Institute of Technology, University of Minnesota, 1972, pp.
457-472.
98. Pitts, G.L.: Augmented Block Guidance for Short-Haul Transportation Systems. Johns Hopkins University Applied Physics Laboratory Report APL JHU CP 019/TPR 023, September 1972, 178 pp. (PB 214 391).
99. Pitts, G.L., and Hinman, E. Augmented block guidance - a control concept for automated
urban transit, APL Tech. Dig., 12,1 (1973), pp. 2-10.
100. Pitts, G.L.: A collision avoidance control law for fixed block, short-headway transportation
systems. AIAA Paper 71-942, Guidance, Control and Flight Mechanics Conference,
Hempstead, NY, August 1971.
101. Whitney, D.E., and Tomizuka, M.: Normal and emergency control of a string of vehicles by
fixed reference sampled - data control. Personal Rapid Transit, Institute of Technology, University of Minnesota, 1972, pp. 383-404.
102. Smith, P.G.: Discrete-time longitudinal control of dual mode automobiles. ASME Paper 73ICT-77. 12 pp.
103.' Hinman, E . J . . ( E ~ :Command
~
and control studies for personal rapid transit, program status,
1974. Johns Hopkins University, Applied Physics Laboratory Report APLIJHU CP 0391
TPR 030, April 1975, 178 pp.
104. Garrard, W.L.: Longitudinal control for automated guideway transit vehicles. University of
Minnesota, Report UMTA-MN-I 1-0002-77-2, July 1977, 107 pp.
105. Fenton, R.E., and Chu, P.M.: On vehicle automatic longitudinal control. Trans. Sci. 11,l
(1977), pp. 73-91.
106. Wilkie, D.: Moving cell control scheme for automated transportation systems. Trans. Sci. 4
(1970), pp. 347-364.
107. Fling, R.B., and Olson, C.L.: An integrated concept for propulsion, braking, control and
switching of vehicles operating at close headways. Personal Rapid Transit. Institute of
Technology, University of Minnesota, 1972, pp. 361-382.
108. Garrard, W.L., and Kornhauser, A.: Design of optimal feedback systems for longitudinal control of automated transit vehicles. Trans. Res. 7.2 (1973), pp. 125-144.
109. Garrard, W.L., and Kornhauser. A,: Use of state observers in the optimal feedback control of
automated transit vehicles. J. Dyn. Sys. Meas. Contr. 95,2 (1973), pp. 220-227.
1 10. Kornhauser, A,, et al.: Optimal sample-data control of PRT vehicles. Personal Rapid Transit
11. Department of Audio Visual Extension, University of Minnesota, 1974, pp. 359-366.
1I I. Yang, S.C., and Garrard, W.L.: A low sensitivity modem approach to the longitudinal control
of automated transit vehicles. J. Dyn. Sys. Meas. Contr. 96,2 (1974), pp. 218-228.
112. Evans, R.R., and Whitten, R.P.: Analysis and demonstration of position error headway protection for PRT systems. Personal Rapid Transit [I. Department of Audio Visual Extension, University of Minnesota, 1974, pp. 335-348.
113. Whitten, R.P., and Wright, R.D.: Analysis of short ramps for dual-mode and PRT stations.
Paper presented at First International Conference on Dual-Mode Transportation,
Washington, D.C., May 1974, Transportation Research Board.
1 14. Whitten, R.P.: Short station ramps for AGT systems. IEEE Trans. Veh. Tech. VT-28,l (1979),
pp. 63-70.
115. Brown, S.J.: Design considerations for vehicle state control by the point-follower method. Personal Rapid Transit 11. Department of Audio Visual Extension, University of Minnesota, 1974, pp. 381-390.
116. Bender, J., and Fenton, R.: On the flow capacity of automated highways. Trans. Sci. 4 (1970),
pp. 52-63.
117. Fenton, R.E., et al.: One approach to highway automation. IEEE Proc. 56,4 (1968), pp. 556566.
1 18. Bender, J., Fenton, R., and Olson, K.: An experimental study of vehicle automatic longitudinal
control. IEEE Trans. Veh. Tech. VT-20,4 (1971), pp. 1 14-123.
119. Garrard, W.L.: Longitudinal control for automated guideway transit vehicles. University of
Minnesota, Report UMTA-MN-I 1-0002-77-2, July 1977, 107 pp.
120. Caudill, R.J., and Garrard, W.L.: Vehicle follower longitudinal control for automated guide-
59 1
592
STEVEN E. SHLADOVER
141. Chiu, H.Y., Stupp, G.B., and Brown, S.J.: Vehicle-follower controls for short headway AGT
systems - functional analysis and conceptual designs. Johns Hopkins University Applied
Physics Laboratory Report APL/JHU CP OSI/TPR 035, December 1976, 109 pp.
142. Chiu, H., Stupp, G.B., and Brown, S.J.: Vehicle-followercontrol with variable-gains for shortheadway automated guideway transit systems. J. Dyn. Sys. Meas. Contr. 99,3 (1977), pp.
183-189.
143. Chiu, H.Y.: A state-constrained vehicle-followerapproach to the station-egress problem. IEEE
Trans. Veh. Tech. VT-28,l (1979), pp. 70-79.
144. Miiller, R., and Nocker, G.: Intelligent cruise control with fuzzy logic. Proc. Intelligent Vehicles 92 Symposium, IEEE, Detroit, MI, June-July, 1992, pp. 173-178.
145. Fujita, Y., Akuzawa, K., and Sato, M.: Radar brake system. Proc. International Symposium
on Advanced Vehicle Control 1994, Tsukuba, Japan, October, 1994, pp. 390-395.
146. Yatabe, T., Hirose, T., Tsugawa, S., and Murata, S.: Control of autonomous vehicles with
vehicle-to-vehicle communication. Proc. CCCT 89, JFAC Symposium, Paris, 1989, pp.
553-558.
147. Tsugawa, S., and Murata, S.: Velocity control for vehicle following through vehicle/vehicle
communication. Proc. 22nd International Symposium on Automotive Technology and
Automation, Florence, Italy, May 1990, pp. 343-350.
148. Pue, A.J.: A state-constrained approach to vehicle-follower control for short-headway ACT
systems. Johns Hopkins University Applied Physics Laboratory Report APL/JHU CP
058/TPR 038, August 1977, 128 pp.
149. Pue, A.J.: A state constrained approach to vehicle follower control for short headway automated transit systems. Proc. 1977 Joint Automatic Control Conference, San Francisco, June 1977, Vol. 1, pp. 401-407.
150. Pue, A.J.: A state-constrained approach to vehicle-follower control for short headway automated transit systems. J. Dyn. Sys. Meas. Contr. 100,4 (1978), pp. 291-297.
151. Pue, A.J.: Implementation trade-offs for a short-headway vehicle - follower automated transit
system. IEEE Trans. Veh. Tech. VT-28,l (1979), pp. 46-55.
152. Shladover, S.E.: Longitudinal control of automated guideway transit vehicles within platoons.
J. Dyn. Sys. Meas. Contr. 100,4 (1978), pp. 302-310.
153. Shladover, S.E.: Longitudinal control of automotive vehicles in close-formation platoons. J.
Dyn. Sys. Meas. Contr. 113 (1991), pp. 231-241.
154. Sheikholeslam, S., and Desoer, C.A.: Longitudinal control of a platoon of vehicles. Proc. 1990
American Control Conference, San Diego, May 1990, pp. 291-296.
155. McMahon, D.H., Hedrick, J.K., and Shladover, S.E.: Vehicle modeling and control for automated highway systems. Proc. 1990 American Control Conference, San Diego, May
1990, pp. 297-303.
156. Hedrick, J.K., McMahon, D., Narendran, V., and Swaroop, D.: Longitudinal vehicle controller design for IVHS systems. Proc. 1991 American Control Conference, Boston, MA,
June 1991, pp. 3107-3112.
157. Hedrick, J.K., and Swaroop, D.: Dynamic coupling invehicles under automaticcontrol. Shen, 2.
(Ed.): The Dynamics of Vehicles on Roads and on Tracks. Proc. 13th IAVSD Symposium,
Chengdu, Sichuan, P.R. China, August 1993. Swets& Zeitlinger, Lisse, 1994, pp. 209-220.
158. Sheikholeslam, S., and Desoer, C.A.: Longitudinal control of a platoon of vehicles with no
communication of lead vehicle information. Proc. 1991 American Control Conference,
Boston, MA, June 1991, pp. 31 02-3106.
159. Sheikholeslam, S., and Desoer, C.A.: A system level study of the longitudinal control of a platoon of vehicles. J. Dyn. Sys. Meas. Contr. 114,2 (1992), pp. 286-292.
160. Chang, K.S., Li, W., Devlin, P., Shaikhbahai, A,, Varaiya, P., Hedrick, J.K., McMahon, D.,
Narendran, V., Swaroop, D., and Olds, J.: Experimentation with a vehicle platoon control system. Proc. Vehicle Navigation and Information Systems Conference, Dearborn,
MI, October, 1991, pp. 1117-1124.
161. Choi, S., and Hedrick, J.K.: Comparison of engine models and distance sensors for vehicle
longitudinal control: analysis and test results in PATH. ITS America Annual Meeting,
Washington, D.C., March 1995.
162. Choi, S., and Hedrick, J.K.: Vehicle longitudinal control using an adaptive observer for automated highway systems. Proc. 1995 American Control Conference, Seattle, WA, June
1995.
593
163. Maciuca, D.B., Gerdes, J.C., and Hedrick, J.K.: Automatic braking control for IVHS. Proc.
International Symposium on Advanced Vehicle Control, Tsukuba, Japan, October
1994, pp. 396-401.
164. Ren, W., and Green, D.: Continuous platooning: a new evolutionary and operating concept for
automated highway systems. Proc. 1994 American Control Conference, Baltimore, MD,
June 1994, pp. 21-25.
165. Athans, M., Levine, W., and Levis, A,: On the optimal and suboptimal position and velocity
control of a string of high-speed moving trains. M.I.T. Electronic Systems Laboratory
Report PB 173 640, November 1966, 71 pp.
166. Garrard, W.L., Hand, G.R., and Raemer, R.: Suboptimal feedback control of a string of vehicles moving in a single guideway. Trans. Res. 6 (1972), pp. 197-210.
167. Levis, A,, and Athans, M.: Sampled-data control of high-speed trains. M.I.T. Electronic Systems Laboratory Report ESL-R-339, PB 177 669, January 1968, 50 pp.
168. Powner, E.T., Anderson, J., and Qualtrough, G.: Optimal digital control of cascaded vehicles
in high-speed transportation systems. Trans. Res. 3,1 (1969), pp. 101-1 13.
169. Peppard, L.E., and Gourishankar, V.: An optimal automatic car-following system. IEEE
Trans. Veh. Tech. VT-21,2 (1972), pp. 67-73.
170. Yang, Y.-T., and Tongue, B.H.: A new control approach for platoon operations during vehicle
exitlentry. Veh. Sys. Dyn. (to appear, 1995).
171. Sklar, S.J., Bevans, J.P., and Stein, G.: "Safe-approach" vehicle-followercontrol. IEEE Trans.
Veh. Tech. VT-28,l (1979), pp. 56-62.
172. Caudill, R.J., and Blasnik, S.L.: Model-reference longitudinal control for automated vehicle
systems. ASME J. Dyn. Sys. Meas. Contr. 102, (1980).
173. Caudill, R.J., Di Matteo, P., and Thomas, S.P.: Longitudinal control for automated highway
vehicles. J. Dyn. Sys. Meas. Contr. 104,2 (1982), pp. 180-187.
174. Youcef-Toumi, K., Sasage, Y., Ardini, J., and Huang, S.-Y.: The application of time delay
control to an intelligent cruise control system. Proc. 1992 American Control Conference, Chicago, IL, June 1992, pp. 1743-1747.
175. Godbole, D.N., and Lygeros, J.: Longitudinal control of the lead car of a platoon.: Proc. 1994
American Control Conference, Baltimore, MD, June 1994, pp. 398-402.
176. Grimes, D.M., and Grimes, C.A.: Cradar - an open-loop extended-monopulse automotive
radar. IEEE Trans. Veh. Tech. 38,3 (1989), pp. 123-131.
177. Alvisi, M., Deloof, P., Linss, W., Preti, G., and Rolland, A.: Anticollision radar: state of the
art. Proc. Advanced Telematics in Road Transport, Vol. 11, DRIVE Conference, Brussels, February 1991, pp. 943-961.
178. Stove, A.G., and Chodyniecki, W.: Radar sensor for AICC. Seminar 9 - PROMETHEUS Future Systems, Autotech '93, Institute of Mechanical Engineering, 1993.
179. Kikuchi, H., Ishiyama, M., and Nakajima, T.: Development of laser radar for radar brake system. Proc. International Symposium on Advanced Vehicle Control 1994, Tsukuba,
Japan, October, 1994, pp. 385-389.
180. Swaroop, D., Hedrick, J.K., Chien, C.C., and Ioannou, P.: A comparison of spacing and headway control laws for automatically controlled vehicles. Veh. Sys. Dyn. 23,s (1994), pp.
597-625.
181. Sekine, H., Tamura, K., Asanuma, N., and Sato, M.: A study of electronic cornering speed
control system. Proc. International Symposium on Advanced Vehicle Control 1994, Tsukuba, Japan, October, 1994, pp. 455-460.
182. Butsuen, T., Doi, A., and Sasaki, H.: Development of a collision avoidance system with automatic brake control. Proc. First World Congress on Applications of Transport Telematics and Intelligent Vehicle-Highway Systems, Paris, November-December 1994, pp.
2079-2086.
183.,Naab, K., and Reichart, G.: Driver assistance systems for lateral and longitudinal vehicle guidance - heading control and active cruise support. Proc. International Symposium on
Advanced Vehicle Control 1994, Tsukuba, Japan, October 1994, pp. 449-454.
184. Reichart, G., and Naab, K.: Systems for lateral and longitudinal vehicle guidance. Proc. First
World Congress on Applications of Transport Telematics and Intelligent Vehicle-Highway Systems, Paris, November-December 1994, pp. 2126-2133.
185. Andrews, M., Jacobi, S., Reuter, U., and Salzer, V.: Porche's contribution to safer and more
comfortable driving in the future. Proc. First World Congress on Applications of Trans-
STEVEN E. SHLADOVER
port Telematics and Intelligent Vehicle-Highway Systems, Paris, November-December
1994, pp. 1879-1886.
186. Kehtarnavaz, N., Lee, J.S., and Griswold, N.C.: Vision-based convoy following by recursive
filtering. Proc. 1990 American Control Conference, San Diego, May 1990, pp. 268-273.
187. Sheikholeslam, S., and Desoer, C.A.: Combined longitudinal and lateral control of a platoon
of vehicles. Proc. 1992 American Control Conference, Chicago, IL, June 1992, pp. 17631767.
188. Pham, H., Hedrick, K., and Tomizuka, M.: Autonomous steering and cruise control of automobiles via sliding mode control. Proc. International Symposium on Advanced Vehicle
Control 1994, Tsukuba, Japan, October 1994, pp. 444-448.
189. Pomerleau, D., Thorpe, C., Longer, D., Rosenblatt, J.K., and Sukthankar, R.: AVCS research
at Carnegie Mellon University. Proc. IVHS America 1994 Annual Meeting, Vol. I,
Atlanta, Georgia, April 1994, pp. 257-261.
190. Okuno, A., Kutami, A,, and Fujita, K.: Towards autonomous cruising on highways. Automated Highway/Intelligent Vehicle Systems: Technology and Socioeconomic Aspects.
SAE Publication SP-833, August 1990, pp. 7-16.
191. Morita, T., Takahashi, H., Satonobu, J., Kojima, K., and Morimoto, Y.: An approach to the
intelligent vehicle. Proc. Intelligent Vehicles 93 Symposium, Tokyo, Japan, July 1993, pp.
426-43 1.
192. Hayafune, K., and Yoshida, H.: Control method of autonomous vehicle considering compatibility of riding comfort and vehicle controllability. Automated Highway/Intelligent
Vehicle Systems: Technology and Socioeconomic Aspects. SAE Publication SP-833,
August 1990, pp. 35-40.
193. Ohnishi, K., Komura, J., and Ishibashi, T.: Development of automatic driving system on
rough road - realization of high reliable automatic driving system. Shigematu, T., Hashimoto, Y., and Watanabe, T.: Development of automatic driving system on rough road automatic steering control by fuzzy algorithm. Ooka, N., Tsuboi, T., and Oka, H.:
Development of automatic driving system on rough road - fault tolerant structure for
electronic controller. Proc. Intelligent Vehicles 92 Symposium, IEEE, Detroit, MI,
June-July 1992, pp. 148-165.
194. Suzuki, T., and Tachibana, A.: A fully automated intelligent vehicle system based on computer
vision for automated highway system. Proc. First World Congress on Applications of
Transport Telematics and Intelligent Vehicle-Highway Systems, Paris, NovemberDecember 1994, pp. 1925-1 932.
195. Ulmer, B.: Autonomous automated driving in real traffic. Proc. First World Congress on
Applications of Transport Telematics and Intelligent Vehicle-Highway Systems, Paris,
November-December 1994, pp. 21 18-2125.
196. Hsu, A., Sachs, S., Eskafi, F., and Varaiya, P.: The design of platoon maneuvers for IVHS.
Proc. 1991 American Control Conference, Boston, MA, June 1991, pp. 2545-2550.
197. Varaiya, P.: Smart cars on smart roads: Problems of control. IEEE Trans. on Auto. Contr.
AC-38,2 (1993), pp. 195-207.
198. Lygeros, J. and Godbole, D.N.: An interface between continuous and discrete-event controllers for vehicle automation. Proc. 1994 American Control Conference, Baltimore,
MD, June 1994, pp. 801-805.
199. Rao, B.S.Y., Varaiya, P., and Eskafi, F.: Investigations into achievable capacities and
stream stability with coordinated intelligent vehicles. Trans. Res. Rec. 1408 (1993),
pp. 27-35.
200. Rao, B.S.Y., and Varaiya, P.: Flow benefits of autonomous intelligent cruise control in mixed
manual and automated traffic. Trans. Res. Rec. 1408 (1993), pp. 36-43.
201. Niehaus, A,, and Stengel, R.F.: An expert system for automated highway driving. Proc. 1990
American Control Conference, San Diego, CA, May 1990, pp. 274-280.
202. Niehaus, A., and Stengel, R.F.: Rule-based guidance for vehicle highway driving in the presence of uncertainty. Proc. 1991 American Control Conference, Boston, MA, June
1991, pp. 3119-3124.
203. Niehaus, A., and Stengel, R.F.: Probability-based decision making for automated highway
driving. Proc. Vehicle Navigation and Information Systems Conference, Dearborn,
MI, October, 1991, pp. 1125-1136.
204. Pilutti, T., Raschke, U., and Koren, Y.: Computerized defensive driving rules for highway
205.
206.
207.
208.
209.
210.
595
maneuvers. Proc. 1990 American Control Conference, San Diego, CA, May 1990, pp.
809-81 1.
Karaaslan, U., Varaiya, P., and Walrand, J.: Two proposals to improve freeway traffic flow.
Proc. 1991 American Control Conference, Boston, MA, June 1991, pp. 2539-2544.
Rao, B.S.Y., and Varaiya, P.: Potential benefits of roadside intelligence for flow control in an
IVHS. Proc. 1994 American Control Conference, Baltimore, MD, June 1994, pp. 418422.
Rao, B.S.Y., and Varaiya, P.: Roadside intelligence for flow control in an intelligent vehicle
and highway system. Trans. Res. C: Emerg. Tech. 2C.1 (1994), pp. 49-72.
Ramaswamy, D., Medanic, J.V., Perkins, W.R., and Benekohal, R.: Lane assignment on an
automated highway. Proc. 1994 American Control Conference, Baltimore, MD, June
1994, pp. 413-417.
Peng, H.: Highway-level vehicle control for AHS. IVHS J. (to appear 1995).
Tsao, H.-S. J.: A functional architecture for automated highway traffic planning. Proc. IVHS
America 1994 Annual Meeting, Vol. 2, Atlanta, Georgia, April 1994, pp. 597-605.
PREFACE
446
PREFACE
and analysing the speed dependency of steering torque characteristics. Using a validated
detailed nonlinear finite-element model, the authors identify turn-slip-related kinematics as the main mechanism for the speed-dependent tyre steering torque. The important
comfort-related aspects of driving noise and vibrations are discussed by Gallrein et al. [3]
by incorporating a structural tyre model into a frequency-based simulation framework.
The authors show that the approach correctly captures tyre behaviour for frequencies of
up to 250 Hz resulting from geometric road excitations. The fourth paper by Schmeitz and
Alirezaei [4] also studies tyre vibrations, but with the view of extracting information on the
tyreroad friction condition. In particular, the resonance peak in the wheel speed signal
excited by road roughness is used to estimate changes in road friction. Results show that
the proposed estimation technique is reasonably accurate. Friction is the main topic of
the fifth contribution by Sharp et al.[5] This study discusses the mechanics of rubber friction by revisiting the classical experiments of Grosch and comparing the measurements
with friction coefficient predictions from Perssons deformation theory. It is shown that
there is value in the current rubber theory but considerable scope for further research on
the mechanics of rubber friction exists. The sixth paper, written by Maniowski [6], closes
the special issue by highlighting the importance of tyres in relation to vehicle handling.
The author explores the influence of temperature-sensitive tyre characteristics on vehicle
behaviour and driving strategy by solving the multi-parametric optimisation problem of
the minimum lap time at a race circuit.
All together, we hope that the reader will find this special issue as interesting as we do
and the new insights and inputs will be useful for researchers and engineers working with
tyres.
We thank the International Technical Committee, the Session Chairs, all participants
and authors who worked hard to help make the 4th Tyre Colloquium a success and facilitated a very interesting programme that allowed a sharing of ideas between experts from
around the world. Also, we are very grateful to the International Association for Vehicle System Dynamics board and, most notably the president Professor Hans True, for the
prompt and full support for organising the Tyre Colloquium at the University of Surrey.
Special thanks go to Professor Manfred Plchl and Jan Prins who were essential in making
this event possible in the first place.
International Technical Committee
Besselink, I., TU Eindhoven, Netherlands
Blanco-Hague, O., Michelin, France
Blundell, M., Coventry University, UK
Gipser, M., Hochschule Esslingen, Germany
Klppel, M., DIK, Germany
Lorenz, B., Forschungszentrum Jlich, Germany
Lugner, P., TU Wien, Austria
Mavros, G., Loughborough, UK
Oertel, Ch., FH Brandenburg, Germany
Pacejka, H. B., Netherlands
Persson, B., Forschungszentrum Jlich, Germany
Plchl, M., TU Wien, Austria
Prins, J., Jaguar Land Rover, UK
447
Special thanks to the sponsors of the 4th Tyre Colloquium (in alphabetical order)
IPG Automotive GmbH, Jaguar Land Rover, MSC Software, MTS Systems Ltd, SIMPACK.
References
[1] Gipser M. FTire and puzzling tyre physics: teacher, not student. Vehicle Syst Dyn.
2016;54(4):448462. doi:10.1080/00423114.2015.1117116
[2] Wei Y, Oertel Ch, Liu Y, Li X. A theoretical model of speed-dependent steering torque for
rolling tyres. Vehicle Syst Dyn. 2016;54(4):463473. doi:10.1080/00423114.2015.1111391
[3] Gallrein A, Bcker M, Roller M. NVH model of a rotating tire. Vehicle Syst Dyn.
2016;54(4):474491. doi:10.1080/00423114.2016.1158844
[4] Schmeitz AJC, Alirezaei M. Analysis of wheel speed vibrations for road friction classification.
Vehicle Syst Dyn. 2016;54(4):492509. doi:10.1080/00423114.2015.1131307
[5] Sharp RS, Gruber P, Fina E. Circuit racing, track texture, temperature and rubber friction.
Vehicle Syst Dyn. 2016;54(4):510525. doi:10.1080/00423114.2015.1131308
[6] Maniowski M. Optimisation of driver actions in RWD race car including tyre thermodynamics.
Vehicle Syst Dyn. 2016;54(4):526544. doi:10.1080/00423114.2016.1158411
1.
Introduction
There are numerous important requirements that a typical automotive suspension must satisfy
[1,2] including
maintaining proper vehicle posture when subject to various inertial and external forces and
moments caused by braking, turning, wind gust, and other events;
providing ride comfort in view of the road roughness inputs which act as a major
disturbance to the vehicle;
securing good road handling and overall vehicle agility;
avoiding excessive suspension stroke or related hard stop/impact.
Additional opportunities exist with the more advanced active and semi-active suspensions
regarding various safety functions [3] and other new exciting functionalities.[4] Although
most current suspensions are of the conventional passive type, this survey will focus on active
suspensions where additional power sources such as pumps and compressors are needed
to accomplish the desired functionality (for a more precise definition of passive and active
*Corresponding author. Email: htseng@ford.com
c 2015 Taylor & Francis
1035
2.
As stated in the introduction and as established from many suspension studies,[7,911] the
most significant and insightful conclusions for vehicle suspensions can be observed from a
simple quarter car model with an undamped tyre and a road disturbance described by Gaussian white noise ground velocities or, equivalently,[7] a step in road displacement (or an
impulse in road velocity). Therefore, in this paper, we will focus our discussion on control
analysis and comparison using the quarter-car model.
The performance indices relating ride comfort and handling are often measured by the rootmean-square (RMS) values of sprung mass acceleration and tyre deflection, respectively.[7]
Additionally, the RMS value of suspension displacement can be utilised to enforce the rattle
space (or available suspension displacement) constraint. As a result, the optimisation of the
suspension controller using the quarter-car model often lends itself to a well-defined H2 or
linear quadratic (LQ) optimisation problem.
1036
2.1.
Performance index
The ground vehicle suspension design is influenced by a number of often conflicting factors
and attributes. These factors will be discussed and then combined into an overall performance
index for the H2 optimisation problem.
2.1.1.
One of the most popular and fitting ride measures is based on the RMS value of vertical
acceleration of the sprung mass, typically measured or projected at the drivers or passengers
seat locations. In a field study in 1978,[12] the authors involved 78 passengers in two different
vehicles and 18 different road sections to conclude that excellent correlation was found to
exist between the subjective ride ratings and simple RMS acceleration measurement at either
the vehicle floorboard or the passenger/seat interface.
Further refinements of the RMS ride measure are possible through adding the RMS value
of the sprung mass jerk (the derivative of acceleration) to the RMS value of sprung mass
acceleration. Some literature [1315] advocates the inclusion of jerk as an added measure
that amplifies the contribution from high-frequency disturbances, which are important for
NVH (noise, vibration, and harshness).
To account for the frequency dependency of human sensitivity to vibrations and the length
of time of human exposure, a standard has been developed by the International Organization for Standardization as ISO 2631.[16,17] It noted that the region of greatest human
sensitivity to vertical vibration lies between 4 and 8 Hz, which roughly includes the various
resonances of human internal organs. The ISO-based metric can be easily incorporated into
the H2 optimisation formulation.
National Aeronautics and Space Administration (NASA) developed a more comprehensive measure that takes into account interdependency of various modes/direction
of vibrations.[18] A similar measure has been adapted and extended to automotive
applications,[19] considering the interdependence between heave, pitch, roll, and other
factors.
The ride metric of choice will depend on the context of its usage. While the more complex ones can reflect more details and nuances, the simple one can better focus on the
system level and major benefits. This paper will utilise the simple RMS of sprung mass
acceleration.
2.1.2.
In practice, the available suspension displacement or so-called rattle space is limited. One can
include this constraint by adding the suspension displacement in the performance index. With
the attempt to minimise the corresponding cost, the H2 optimisation problem effectively soft
constrains the suspension displacement. Similar to the ride comfort metric, the suspension
displacement-related cost can be reflected on the RMS value of suspension displacement,
especially for random/stochastic road input.
2.1.3.
As there exists a convex relationship between tyre cornering forces and normal forces
(reflected by tyre deflection), the variation of tyre deflection leads to the variation of cornering forces. It is conceivable that a very large variation of tyre deflection may even lead to
1037
a loss of contact with the ground. Therefore, the RMS value of tyre deflection variation can
be an explicit metric for handling characteristics. Note the direct relationship between tyre
deflection and cornering capabilities have been experimentally studied.[7]
Based on the aforementioned discussion, it is no surprise that the most widely used performance index for active suspension study is the combination of the three previously discussed
RMS values. That is, using a weighted combination of RMS acceleration of sprung mass,
RMS suspension displacement, and RMS tyre deflection variation to represent the performance index for optimisation. This performance index lends itself to an H2 norm of the
weighted states and input of the quarter-car model. Interestingly, the deterministic and the
stochastic interpretations of this H2 norm lend themselves to the road disturbance of a step in
road displacement, and that of Gaussian white noise ground velocities, correspondingly. The
equivalence of these two interpretations was described in more detail in [20].
2.2.
The mathematical state space representation of a 2 DOF quarter-car model (Figure 1) can be
found in [9]. Here, we choose the states as tyre deflection, unsprung mass velocity, suspension deflection, and sprung mass velocity (Figure 1). With this representation, we have the
performance index as
PI = E(r1 x21 + r2 x23 + u2 ),
(1)
where x1 and x3 represent the tyre deflection and suspension stroke as illustrated in
Figure 1(a), u represents the sprung mass acceleration (u = U/ms ), and the expectation E
represents steady-state mean-square values. Figure 1(b) shows the corresponding limiting
case when unsprung mass is negligible, and Figure 1(c) illustrates a passive quarter car.
Figure 2 shows a quarter car equipped with various suspension elements including
adjustable spring and damper and different configurations such as springs in parallel to the
actuator and in series to the actuator.
The earlier four-state LQ optimisation was solved numerically,[7,9] and the results of
the corresponding global study for various weighting in the combination of performance
index are shown in Figures 3 and 4. The figures illustrate how the weighting factors serve as
tuning knobs, controlling the trade-off between acceleration, suspension stroke (design constraint), and tyre deflection (handling performance constraint). These plots were generated
with the ratio of sprung-to-unsprung mass set to 10 and the unsprung natural frequency (or
wheel-hop) as 10 Hz. The choice in optimisation with respect to different combinations of
Figure 1. (a) Two DOF quarter-car model, its states x, road velocity w, and actuator input U; (b) one DOF
quarter-car model; and (c) two DOF quarter-car model with passive spring and damper.
1038
Figure 2. (a) Two DOF quarter-car model equipped with actuator, serial spring, and parallel adjustable spring and
damper and (b) the corresponding bond graph model.
weighting in the performance index can be predetermined as the vehicle design suspension
characteristic or adapted dynamically during on-road operation to enable the best driving
experience.
We see from Figures 3 and 4 that reducing r1 , the weighting factor on tyre deflection,
while maintaining r2 , the weighting factor on suspension deflection, results in increased tyre
deflection and decreased sprung mass acceleration, that is, better ride at the expense of somewhat worse handling. On the other hand, keeping r1 constant while reducing r2 leads to
increased suspension deflection and decreased sprung mass acceleration. In the extreme case
of assigning r2 to zero, the problem becomes a degenerative one where extremely large suspension displacement would result. The corresponding optimal force of this degenerated case
would not attempt to support the static weight of the vehicle since there is no cost associated
with suspension stroke. As a rule of thumb, choosing r1 /r2 to about 10 leads to reasonable
results.[7] The shaded area in Figures 3 and 4 illustrates where most optimal designs for
automotive applications will reside due to the above-mentioned practical considerations. An
important observation is that no suspension control design, linear or nonlinear, can exhibit
behaviour in the area falling below the r1 near zero line (represented by r1 = 102 ) or the r2
near zero line (represented by r2 = 103 ).
Similar solutions can be obtained using linear matrix inequality (LMI)-based
approaches.[2123] A desirable feature of the LMI-type controllers is that they can incorporate the slowly time-varying parameters, such as changes in vehicle sprung mass. Related
methods here include the linear parameter-varying techniques [22,24] that explicitly include
possible slowly varying system parameters, such as road roughness and vehicle speed
variations with results similar to gain scheduling investigated in [25].
Various observations about the suspension performance limitation from different
approaches as well as limited experimental work [7] have affirmed the performance limits
observed in Figures 3 and 4. For example, any improvement in one area (e.g. ride) would be
reflected through corresponding loss of performance in other complementary areas (e.g. handling in wheel-hop mode and/or rattle space requirement). Essentially this is a consequence
of the fact that a single actuator is required to fulfil two conflicting requirements. This is also
reflected through the invariant properties reported in [26].
1039
Figure 3. Normalised ride comfort versus suspension stroke (A as road roughness coefficient and V as vehicle
speed).
2.3.
Illustrative example
To illustrate the performance in frequency response as well as where it resides on the global
carpet plot of Figures 3 and 4, a practical example of H2 optimal design active suspension
is described in the following.
Assume that for a particular vehicle = 10 and us = 2 10 rad/s. The road is described
by a road roughness coefficient A = 1.6 105 ft (4.9 106 m) which corresponds to a
medium quality road. The vehicle is traversing this road at a speed of V = 80 ft/s (88.5
km/h). It is reasonable and desirable to design a suspension such that the tyre deflection from
equilibrium under this road excitation would be maintained within 1 inch (2.54 cm) 99.7%
of the time. This would translate to a requirement on the normalised tyre deflection, x1 , to
RMS
be maintained within 0.3 s1/2 . Note that we use the accent to denote the normalised
value with respect to the road velocity excitation, for example, x1 := x1,rms / 2 AV , where
A is the road roughness coefficient and V is the vehicle travelling speed.[27]
From Figure 4 and a normalised RMS tyre deflection value of 0.3 (s1/2 ) in this illustrative
example, we see the minimum possible normalised RMS acceleration u is 10 (s3/2 ). If we
accept the normalised value of 10.9 (which reflects to a 0.03 g sprung mass acceleration RMS,
1040
or the sprung mass acceleration under 0.1 g 99.7% of the time), we can arrive at this optimal
control law by picking the tuning factors r1 = 1100 and r2 = 100. This is represented as
design point A in Figures 3 and 4. Using these values for r1 and r2 in Figure 3 reveals a
normalised RMS secondary suspension deflection x3,rms = 0.605 s1/2 which ensures that the
suspension deflection will remain within 2 inch (5.08 cm) of the static value 99.7% of
the time. With r1 = 1000 and r2 = 100, the optimal control gains for full state feedback are
k 1 = 6.084, k 2 = 0.548, k 3 = 10.0, and k 4 = 4.438 for tyre deflection x1 , unsprung
mass velocity x2 , suspension deflection x3 , and sprung mass velocity x4 , respectively. The
closed-loop eigenvalues are e1,2 = 2.20 j2.26 and e3,4 = 2.75 j62.9. The first set
of oscillatory eigenvalues corresponds to the vehicle heave mode with a natural frequency
of 0.5 Hz and damping ratio of 0.7. The second set corresponds to the wheel-hop mode
with a resonant frequency of 10 Hz and relatively small damping ratio of 4.4%. Whether or
not this small amount of damping is of consequence will depend on the particular hardware
implementation, which may include a combination of active and passive means to improve
closed-loop robustness. Also in practice, an excessively soft suspension setting may require
load levelling to keep the suspension static or steady-state deflection in the neighbourhood of
zero or some nominal value.
1041
Figure 5. Normalised ride comfort versus handling passive and active suspensions.
At this stage, it is instructive to compare the above active suspension results with a corresponding passive case with the structure shown in Figure 1(c). For this purpose consider
Figure 5 which displays normalised RMS acceleration versus tyre deflection for active and
passive suspensions. For simplicity, the optimal active suspension performance is shown via
the limiting curves r1 0 and r2 0 only, which are representative of the main trends.
Here, the design points with lower accelerations and higher tyre deflections are characterised
by softer, lower frequency body modes and less damped wheel-hop modes.
The added curves in Figure 5 correspond to the passive performance trade-offs with sprung
mass mode resonant frequencies f2 at 1 or 1.5 Hz (which corresponds to different passive
spring stiffness in Figure 1(c)), and damping ratios varying between 0.02 and 1 (which
corresponds to varying passive damping in Figure 1(c)). A typical passive suspension setting
is shown as design point P1 , where f 2 = 1 Hz and = 0.3. As can be seen from Figure 5,
this amount of damping is near optimal in terms of the ride and handling compromise for the
passive case. These values are within the range typically found in most of todays vehicles
1042
which have evolved through many iterations primarily based on intuition and decades of
experience.
The performance of an optimal active suspension with the same amount of wheel-hop
as the optimal passive is shown in Figure 5 as design point A1 . Comparing design points
P1 and A1 , it can be seen that in this case the active suspension results in only 11% lower
RMS acceleration levels with respect to P1 . However, it can also be seen that for some other
road/speed operating conditions (to the right of A1 ), there is a potential for substantial ride
improvement with active suspensions compared with their passive counterparts. Indeed, as
shown in Figure 5, for passive suspensions almost any deviation in operating conditions from
P1 will result in degradation of performance. On the other hand, for active suspensions, either
handling or ride can be improved by choosing the tuning parameters r1 and r2 such that the
resulting operation settles either to the left or to the right of A1 , respectively. For example, if
we pick a different active suspension design point A (Figure 5) by relaxing the tyre deflection
constraints, we see that the RMS acceleration at point A is as much as 67% smaller than that
at Point P1 .
Based on the above results and observations, it can be concluded that the full advantage
of active suspensions stems from possible adaptive tuning (or gain scheduling) of controller
parameters, depending on the driving condition.[7,25] For example, if the steering wheel
position or lateral acceleration sensors (possibly augmented with other road information)
indicate operation on a straight section of a road where handling is less critical, then it may
be possible to relax the wheel-hop constraint. Similarly, by knowing which sections of the
road the vehicle is or will be travelling on and their corresponding road roughness, the tyre
deflection constraints can be enforced differently, allowing more substantial improvement in
RMS acceleration. Note that the exact amount of wheel-hop that could be tolerated under
different driving conditions can be predetermined through appropriate vehicle tests.
The frequency responses of the three metrics with respect to road disturbance velocity offer
insightful information between the trade-offs in suspension design. Different attributes of the
performance at three operating points of interest (A, A1 , and P1 ) are illustrated in Figure 6
in their corresponding Bode plots. From the sprung mass acceleration frequency plot, it can
be seen that the main difference between the active suspension A1 and corresponding passive
counterpart P1 occurs at the sprung mass mode around 1 Hz, where the active suspension
brings much more damping through the damping on sprung mass velocity, the so-called
skyhook damping. The active suspension A results in further significant reduction of the
acceleration levels (except at the invariant point at the wheel-hop frequency [26]), but at
the same time it deteriorates handling and rattle space performance due to a large, poorly
damped resonant peak around the wheel-hop frequency. Note that the invariant point arises
from the structure of the suspension and results in the performance limit of sprung mass
acceleration at the wheel-hop frequency regardless of control design, assuming negligible
tyre damping. The reduction of acceleration at other frequency regions causing the rise of
suspension and tyre deflection can also be interpreted as the waterbed effect, a term coined
in H infinity control analysis, that is, the improvement in one area (e.g. ride with respect to
sprung mass frequency) of the frequency domain would be reflected through corresponding
loss of performance in other complementary areas (e.g. handling in tyre-hop mode and/or
rattle space requirement).
The suspension deflection Bode plot of Figure 6(b) indicates that, unlike their passive
counterparts, the LQ optimal suspensions can result in a non-zero gain in very low frequency
between the rattle space and road velocity input.[7] However, in practice this may not be
detrimental since it is usually a standard practice to use appropriate signal processes, for
example, high-pass filters to eliminate the very low-frequency components of the ground
input signals resulting from hills and extended road grade.
1043
Figure 6. (a) Tyre deflection Bode plot of points on Figure 5 and point A with dynamic absorber, (b) suspension
Bode plot of points on Figure 5 and point A with dynamic absorber; and (c) acceleration Bode plot of points on
Figure 5 and point A with dynamic absorber.
Moreover, Figure 6 shows the Bode plots of an active suspension control (designed with
the same weighting used in operating point A) combined with a dynamic vibration absorber.
We see that the sprung mass acceleration can be further improved in the wheel-hop frequency
since the introduction of the dynamic absorber (which interacts with only the unsprung mass)
presents a new structure.
2.4.
In semi-active suspension systems, the suspension force can be modulated through a very
small amount of supplied energy. In this case the suspension force cannot track an arbitrary desired force from the unconstrained LQ optimisation-derived control law. Rather, the
problem becomes a constrained LQ optimisation. As a compromise, the semi-active control
design typically follows its unconstrained active counterpart when it can, and operates along
the passivity envelope when it cannot. For example, the damping force is adjusted to follow
the desired suspension force derived from the optimal control law, and set to zero when a negative damping force is required. This control is therefore commonly referred to as clipped
optimal. This may result in 20% lower performance in ride comfort when compared to its
active suspension counterpart.[11]
1044
The optimal control law for the semi-active system has been posed as a constrained LQ
optimisation and solved numerically in [28,29], involving the iterative solution of a timevarying force constraint. A specific example [29] shows that a 10% advantage with respect
to clipped optimal can be achieved. However, it also finds that the amount of improvement
depends on driven scenarios and is usually very limited. A later work [30] leveraged the
explicit hybrid MPC to confirm analytically the previously numerical finding that clipped
optimal is not the optimal control for semi-active suspensions in general.
2.5.
There are numerous investigations of the potential performance of the suspension if the road
disturbance profile can be known in advance.[3135] The solution consists of state feedback
terms that have the same gains as the non-preview case, and feedforward terms that consider
the road disturbance of the upcoming known duration.
It is worth noting that most of the papers place the origin of time (i.e. start of preview
time for future disturbances) at the quarter-car location [31,35,36] or at the location of the
front wheels.[32,37,38] In [34], this origin is simply moved to the location of preview,
thus transforming the preview problem into an equivalent delay-time problem, where an
extensive literature base already exists.[39] This approach is also used in [40] where the
authors developed a method for direct and efficient calculation of the related performance
index.
Comprehensive studies of the benefit of preview compared to its non-preview counterpart
have been demonstrated for full-bandwidth active suspensions [34] and limited bandwidth
ones.[41,42] The benefit for full-bandwidth active suspensions was illustrated on a plot similar to Figures 3 and 4 in [34], relating to its non-preview counterpart. And the benefit for
limited bandwidth active suspensions in ride comfort and suspension stroke improvement
was shown in [42].
3.
Although not always explicitly mentioned, most of the studies with traditional simple ride
models (e.g. quarter-car, half-car, and full-car) focus on road roughness disturbances only,
neglecting external (inertia) forces and loads due to, for example, braking and turning. The
control damping for inertia load-induced excitations can be included later in a separate and
independent design procedure. In practice, this is accomplished using longitudinal and lateral
acceleration measurements as feedforward signals [43] indicating braking or turning actions.
This typically leads to appropriate stiffening of the active or semi-active suspension to prevent or counteract excessive pitch or roll. This stiffening action may reduce the ride quality,
but in practice this may be of secondary importance, especially during hard braking and/or
handling manoeuvres.
For the case when complete decoupling of the above road and load disturbance effects is
desired, Smith,[44] Smith and Wang,[45] and Wang and Smith [46] present an interesting
approach based on Youla parameterisation and LFT (linear fractional transformation). The
resulting controller can be designed independently first for the road-induced disturbance
and then for the load counterpart. For example, the suspension can be made very stiff for
load inputs in order to hold a desired attitude, while at the same time it could be made very
soft for the ground inputs (of course, the latter has to be within some practical limits since
any prolonged ramp-like ground input could lead to suspension bottoming and undesirable
1045
jerks and possibly even structural damage). This is somewhat similar to the above-mentioned
sequential design approach used in practice with the exception of total decoupling. In this
context, the practical design uses longitudinal and lateral acceleration signals (in addition
to other heave-related measurements) as a feedforward signal or proxy for gas/brake pedal
and/or steering wheel applications.
A comprehensive and direct approach to simultaneous inclusion of road and load disturbances has been recently presented in [47] based on the above more conventional LQ
stochastic optimisation. In addition to the usual road disturbance, which is assumed unmeasured, Brezas and Smith [47] also introduce the load disturbance but as a measured, constant
disturbance. An advantage of this approach is that it directly addresses optimisation of the
main attribute of interest, that is, ride comfort, which is introduced through an appropriately
formulated LQ performance index.
In the case of longitudinal dynamics, further extensions of this approach are possible by
introducing the force disturbance through a delay, which effectively presents a load preview
from the standpoint of the suspension controller similar to the preview modelling used in
[34]. This could be further refined by using a (simple) model of vehicle power train (brake)
paths, from gas (brake) pedal to actual wheel torque production. Similar comments apply for
load prediction due to steering inputs where ultimately some appropriate models of vehicle
handling can be used too.
A simple approach to load disturbance containment is through load levelling, that is,
applying an integral control to regulate the rattle space between sprung and unsprung
masses.[4851] In particular, Hrovat and Hubbard [48] use jerk-optimal LQ control to end
up with an effective and systematic way of obtaining relatively fast load levelling with
some mostly minor sacrifice in sprung mass acceleration performance. It should be
pointed out that in practice a slow-acting load levelling is often used to centre the suspension against slow-changing loads. Davis and Thompson [49] use both integral and derivative
controls in a ride quality- and handling capability-focused 7 DOF full-car model to regulate
as well as maximise usable suspension deflection as the derivative action helps to decrease
overshoot.
As the suspension stroke is limited, practical design of all suspensions may introduce
additional feedback elements to stiffen up the suspensions as the rattle space approaches
its travel limits. In addition, actuator dynamics may contain delays, nonlinearities (which is
inherent in the semi-active case), or lags. Various control designs have been developed to
address these issues including backstepping design,[52] adaptive robust control,[53] learning automata,[54,55] and Pareto optimisation.[56] In [52], a nonlinear filter whose effective
bandwidth depends on the magnitude of the suspension travel was introduced to facilitate smooth shifting of the control objective between ride comfort and suspension stroke
utilisation. In [53], a two-loop controller with adaptive robust control was introduced to
ensure that the nonlinear actuator delivers the force requested from the LQ top (outer) loop
design. Gordon et al. [54] and Howell et al. [55] used an approach combining concepts from
stochastic optimal control with those of learning automata. And Valasek and Kortum [56]
used the Pareto optimisation procedure to determine the state-dependent feedback gains in a
predefined control structure.
It is worth noting that Gordon and Best [57] included non-quadratic terms for the optimisation of semi-active suspensions in their optimisation and observed the performance with
the conclusion that the standard clipped optimal technique is close to the one resulted
from a complex optimisation. In [58], the optimisation of semi-active suspension with preview is numerically optimised and it was observed that only small improvement can be
achieved compared to the simple clipped optimal preview control. These conclusions are
corroborated in [30] where analytical explicit model predictive control was used.
1046
4.
In addition to the above LQ optimisation-based approaches that are the main focus of the
present survey, there are many other possible approaches to advanced active and semi-active
suspension design and analysis. These include Youla parameterisation and LFT,[4446] pole
placement,[59] LMI,[60] H2/H-infinity, and H-infinity.[6165]
Furthermore, multi-objective optimisation (symbolically or numerically) has been widely
used to systematically manage nonlinearities and constraints.[6668] It is particularly
prevalent for semi-active suspension applications.[69,70] As the suspension performance
index often includes conflicting terms, the solution of a multi-objective optimisation finds
the best trade-off among the various predefined control terms. This is also known as Pareto
optimisation used in systematic design procedures (note that a similar result for the linear
system was obtained using the previously mentioned global optimisation in Section 2 with
varying weights).
While ride comfort is the main focus for passenger vehicle suspensions, minimising road
damage is important for truck and heavy vehicle suspensions.[7174] An approach that pays
particular attention to tyre force variation is an interesting extension of the well-known
skyhook concept,[8] proposed in [75] as the so-called ground-hook control. Specifically, it
amounts to augmenting the (semi-)active actuator desired force calculation with an additional, virtual damping term proportional to the relative velocity between the unsprung mass
and ground (Figure 7), emulating a fictitious damper between the unsprung mass and road.
The typical objective of such a suspension optimisation was to minimise the integral square
of the tyre force [76,77] by using a simple quarter-car vehicle model for the control synthesis, which is similar to previously mentioned LQ-based RMS tyre deflection minimisation
except that the latter would typically be part of a performance index that includes additional
terms due to ride comfort and rattle space (design) constraint. The integral square sprung
mass acceleration term as a measure of ride comfort was later included in [78,79] where
the ground hook and other coefficients were determined from numerical optimisation with
specified cost function. Note that the above optimisation for the case of semi-active actuator
amounted to essentially nonlinear system parameter optimisation since the extended groundhook optimisation structure was fixed a priori.[76,77] Since the main nonlinearity is in the
semi-active actuator, state dependence of the related parameters was also introduced in the
process of optimisation.[56]
The ground-hook concept was found useful when minimising road damage [56,77] especially due to heavy trucks where the damage to pavement can be substantial and up to 10,000
times greater than due to passenger cars.[78] This leads to so-called road-friendly suspension,
Figure 7. (a) Skyhook, (b) ground hook, and (c) combination of skyhook and ground hook.
1047
which was further extended to bridge-friendly [7981] and ESC,[82] that is, Electronic Stability Control-friendly, applications. The concept was further tested through experimental
vehicle implementation [56,77] using a SKODA-LIAZ prototype truck.
The main emphasis in Section 2 was on advanced active and semi-active suspensions
using simplified (minimal component count) system models. In practice, however, any implementation of these advanced concepts will also include additional conventional, passive
components. For example, an active actuator may be positioned in parallel with a spring
for static support and a damper to dissipate energy without an active element. Thus, one must
eventually consider designing the whole suspension system consisting of both active and
passive elements.
In addition, as is typically the practice when implementing the LQ or any other linear
algorithm, one should first attempt (static) linearisation of any possible nonlinearities in actuators, plants, and/or sensors. The combined controller-actuator subsystem should then be
further verified and as necessary refined through additional simulations based on a more
detailed performance-oriented model. The latter should include a detailed model of actuator dynamics, which is often the most challenging and most important component of an
advanced, active, or semi-active vehicle suspension system design.
A reasonable starting point in an active or advanced suspension system design would be
a consideration of the passive system alone in order to establish the best passive benchmark
against which any advanced mixed passiveactive system could then be compared. At this
stage there is a need to more precisely define passivity so that the whole passive suspension
optimisation problem can in turn be properly defined. One of the first attempts at this was in
[28] where the notions of passive operators from mathematics and electrical circuits are used.
In addition, Hrovat [28] also posed the corresponding optimal passive suspension problem.
However, at the time of its writing, much of the necessary mathematical Computer Aided
Engineering (CAE) tools and computer optimisation capability were missing. This class of
problems were effectively treated more than a quarter-century later [83] using LMIs and
other modern CAE tools. In [83] problems with both H2 and Hinf optimisation cost are
considered for the case of a quarter-car model with both ground and load inputs included.
This approach could likely be further extended to hybrid configurations where both passive
and active suspensions are combined.
Once the optimal passive transfer function is determined, there remains the problem of synthesising or actually implementing it via standard passive elements such as springs, dampers,
and inertias. In the process, there is a need for an inertia element called inerter in [84]
that acts on a relative velocity rather than the customary absolute velocity of relevant
masses or inertias. Perhaps the easiest way to see this is via bond graphs where one could
imagine a hydraulic pistoncylinder combination between sprung and unsprung masses [85]
(inerter-like structure was described in its third edition with more examples exhibited in its
later editions). Another example can be found in modelling hydraulic mounts.[86,87] An
effective and practical mechanical implementation of an inerter was shown in [83,84] consisting of multiple gear sets with large gear ratios resulting in a large reflected inertia. The
latter facilitates relatively small overall weight for the device, which is an important practical
consideration. In addition to possible suspension applications, the device was also used to
mitigate steering instabilities of motorcycles.[88]
Another more constrained approach to optimisation was considered in [89] where the
underlying passive structure was first fixed and then the optimisation was performed for each
free design parameter and compared with the best tuning of the conventional passive suspension structure consisting of a parallel spring and damper. The results indicate around 10%
and more improvements in the comfort RMS acceleration metric, although it is not clear if
this included the same level of other constraints (tyre and rattle space).
1048
Using the above-mentioned definition of passivity it is now possible to ask the following question. For a given optimal suspension control law can we determine if it could be
implemented through a passive device? If the answer is positive this would have significant
practical implications since such a device could then be realised using only passive components that are typically less expensive, less complex, and require no external power sources
and consumption to operate. It was first shown in [6,28] that for the quarter-car LQ optimal control the resulting control law is necessarily active. A more comprehensive study was
subsequently undertaken in [5] including additional control laws applied to half- and full-car
cases.
In addition, Smith [44] and Smith and Walker [5] consider some other interesting aspects of
active and passive suspensions. These include the derivation of a complete set of constraints
for the road disturbance transfer functions for typical choices of measured outputs as well
as performance limitations and constraints for active and passive suspensions for quarter-,
half-, and full-car models. This complements the prior well-known results on the invariance
of certain transfer functions at special frequency points.[26]
While the quarter-car model largely describes the vehicle response to the road disturbance,
the heave and pitch experienced at various observed points along the vehicle longitudinal axis
can vary due to the multi-input disturbance (coming through both front and rear wheels). For
passive vehicles, due to the fact that the excitation at the rear wheel is a time-delayed version
of the front wheel, there is a wheelbase filtering effect [90] that changes the transfer function
to heave and/or pitch at particular frequencies as a function of vehicle speed and wheel base.
This effect is more perceptible for trucks with a long wheel base. The quarter-car suspension
design can be expanded to full car. One attractive approach to address the full-car and half-car
suspension design and leverage the extensive knowledge from the quarter-car study is to first
apply control design decoupling modes. The importance of decoupling between sprung mass
pitch and heave mode as well as between the equivalent front and rear quarter-car models
is described in [7,91]. In [91], an active suspension was designed to decouple sprung mass
heave and pitch mode through zeroing the coupling terms between pitch and heave relative
(suspension) deflections, and similarly those between pitch and heave relative (suspension)
velocities. This structurally decoupled active suspension minimises body heave motion to
road pitch disturbance, and similarly body pitch to road heave, to a secondary level. It was
exhibited [91] that a superior performance can be achieved by simply adding the skyhook
(or absolute) damping to the decoupled active suspension design. This approach is further
verified in [92] through mathematical manipulation and state space transformation where the
heave and pitch modes and corresponding disturbances are decoupled.
Important aspects of state estimation are addressed in [93,94]. Analytical techniques for
fault detection and facilitating safe start-up and shutdown of active system are described
in [95]. Some recent attempts to introduce risk sensitivity into the LQ optimal suspension
problem were addressed in [47]. The approach has some practical appeal since it facilitates
tuning the balance between performance and associated risk due to plant uncertainties, modelling limitations, and possible unmodelled dynamics, thus providing system designers with
additional flexibility through an added degree of tuning freedom.
5.
5.1.
Various self-levelling systems have been introduced to the market including the Electronic
Air Suspension systems developed by Citroens hydropneumatic systems [96] and Ford
1049
5.2.
1050
5.3.
A variable geometry active suspension adjusts the ratio of wheel movement to the deflection
of the suspension spring in real time. By changing the leverage of the passive suspension
spring with wheel motion, it essentially controls the wheel rate or effective spring stiffness.
Various systems and hardware configurations that provide variable suspension geometry have been proposed in the literature [107110] including the Delft Active Suspension
concept that was implemented as a prototype vehicle and demonstrated experimentally.[109]
Ideally, one would like to adjust the variable suspension geometry without requiring much
power. One way to do this is to adjust the suspension leverage by changing the suspension
spring mounting point which requires only actuation perpendicular to the (vertical) suspension load or vehicle weight (Figure 8).[108] Ideally, with the perpendicular configuration, the
mechanism would require very low power and low energy. In practice, however, the precisely
perpendicular arrangement would be compromised by suspension motion and deflection.
There may be additional trade-offs in mechanical design forcing some deviation from the
ideal geometry. The trade-offs may include vehicle stance/self-levelling, range of mechanism motion (and resulting range of leverage ratio), packaging, effect of jounce/rebound
motion, and effects on wheel turning. A system design study with the trade-offs in mind
is discussed in [107].
Another form of leverage adjustment concept (Figure 9), the previously mentioned Delft
Active Suspension, is described in [107,111], and its practical realisation with a cone mechanism is illustrated in [112,113]. This mechanism connects the spring to the car body on one
end and to a rotatable crank on the other end. The crank is joint-connected to the suspension/wheel control arm and can be rotated at the joint around the base of the imaginary cone.
The cone mechanism serves two purposes: (1) the length of the spring remains the same as
the crank rotates and (2) the ratio of movement between the wheel/tyre control arm and the
crank changes as the crank rotates (Figure 10).
Due to suspension deflections and motions when experiencing road disturbances, the alignment of the ideal position of the cone mechanism will be affected in practice. In particular,
the base plane of the cone mechanism, that is, the plane of the crank motion, will rotate.
Nevertheless, this is a promising technology since it could provide some active suspension
characteristics with only a semi-active type of actuator.
5.4.
A number of automobile manufacturers have been introducing active and semi-active suspensions that focus on roll control to improve ride and handling with moderate cost and
1051
complexity. It is well known that good ride comfort can be achieved with softly tuned passive or semi-active suspensions combined with soft roll bar settings. However, this will result
in large pitch and roll motions during more aggressive braking and cornering. The active
(or semi-active) roll control is then used to satisfy these conflicting constraints. In practice,
the concept is typically realised by inserting a small hydraulic jack and a gas spring accumulator with a control solenoid, all placed in series with the anti-roll bar. The anti-roll bar
can be as stiff as in street race cars, without negatively impacting straight-line ride since the
solenoid connecting the hydraulic jack and the accumulator is then kept open, thus lowering
the effective stiffness of the roll bar/accumulator combination.
The performance of anti-roll/roll control devices varies as it may be achieved by a mechanical regulator that depends on suspension motion [114,115] or by an electronic controller
that monitors additional non-suspension-related vehicle signals. It can be actuated with a
semi-active device that manipulates hydraulic valves or the like, or with an additional active
1052
device such as a high-capacity pump. An idea similar to active tilt control has also been
explored in [116,117].
Citroen introduced hydropneumatic suspensions in 1955 with its model DS. The hydropneumatic spring-absorber system incorporates the benefits of both the compressibility of
gas (nitrogen) and responsiveness of (incompressible) hydraulic fluid. As the load increases,
it compresses the suspension pistons and causes the gas volume to decrease and pressure to
rise; the gas spring becomes stiffer but keeps the sprung mass frequency roughly the same.
As the fluid passes through a two-way valve, damping is also provided. In addition to its selflevelling feature that regulates ride height and resonance frequency (through stiffer spring
with increased load), the Citroens hydropneumatic system can be applied towards anti-roll
or roll control by manipulating the flow of hydraulic fluid between left and right corners.
Various improvements have been made in Citroens hydropneumatic suspensions. Most
notable are the changes from mechanical regulation to electronic control, the improved capability of the pump, and the associated increase in actuation bandwidth. For the roll control
feature in particular, there have been several incarnations ranging from semi-active (Hydractive I and II) to active (Activa). Citroens semi-active roll control (Hydractive) limits the roll
angle to about 3 in soft mode or 2.5 in hard mode for a 0.6 g constant turn, while its active
roll control (Activa) can eliminate roll angle completely due to the active and electronic
manipulation of hydraulic fluid.[96]
Roll control can be achieved with the help of a hydraulic rotary actuator inserted around
the middle of the stabiliser bar, as stated above, or it can be implemented through a rotary
actuator fitted with the anti-roll bar as in the BMW Dynamic Drive concept.[118] The BMW
system monitors primarily the steering angle and lateral acceleration, among other signals,
and distributes the front and rear axle anti-roll torques. It controls the front and rear actuators
by means of two electronically regulated pressure control valves to minimise the roll motion
in corners and road input roll disturbances, thus improving response to vehicle steering for
vehicle dynamics handling and agility. Note that the roll angle will gradually increase at high
lateral acceleration as the vehicle approaches its stability limit.
It is understood that the benefits of active roll control in overall vehicle stability and active
safety can be further explored through coordinated action with other systems such as active
brakes, active steering, and electronically controlled all-wheel drive. This can be particularly
useful in emergency manoeuvres such as a sudden need for vehicle stabilisation and control
during evasive actions on slippery road surfaces.
5.5.
In general, active suspensions can be categorised into narrow bandwidth and broad bandwidth, referring to the control bandwidth of the actuation systems. This typically refers to
actuator force generation bandwidth for all possible motions across the actuator mounting
points. The narrow bandwidth systems typically exhibit actuator control bandwidth around or
slightly above sprung mass frequency and would use the passive spring, positioned in series to
the actuator, to address high-frequency disturbances and vibrations. As such, their bandwidth
is typically between 3 and 7 Hz.[119] Broad bandwidth systems, on the other hand, attempt
to address oscillations beyond the sprung mass frequency using high-bandwidth actuators
typically above 10 Hz.[119]
Since the broad bandwidth active suspensions (BBASs) attempt to control both the slower
body modes and the fast wheel-hop modes, it is generally expected that the broad bandwidth
systems will be more challenging to implement. It is expected that such suspensions will
incur higher costs, complexities, and possible fuel consumption.[120] However, one should
1053
also observe that narrow bandwidth systems may experience similar fuel consumption due
to either their higher compliance and the corresponding increased pump flow/displacement
requirement [119] or higher control effort associated with the mismatch between the desired
and actual control bandwidth.[43]
5.5.1.
Some narrow bandwidth systems have been brought to the market with limited success. An
example is the Nissan Infinity Q45a [121] whose hardware included: (1) an accumulator (for
bandwidth limitation and NVH containment); (2) a mechanical supporting spring; and (3)
a relatively inexpensive pressure control valve. It had a single-acting actuator configuration
which is simple and inexpensive. It does, however, limit the system controllability on the
rebound stroke, which is in good part determined by the spring/accumulator settings.
Production hydraulic active suspension systems typically use pressure control valves to
regulate the desired force. Considering that a force-generating actuator is accompanied by
corresponding displacements and velocities of the in-series compliance in the total system,
one may consider the flow control valves to regulate the desired displacements/velocities.
Benefits and limitations of both force and displacement control approaches have been
discussed in [119,122124].
DaimlerChrysler introduced the Active Body Control (ABC) system in the 2000 CL
sedan. It consists of four low-bandwidth (5 Hz) hydraulic actuators providing active load
levelling, heave, pitch, and roll control. The ABC suspension system consists of [125,126]
hydraulic cylinders in series with steel coil springs (Figure 11). This cylinder sets the spring
preload and the series arrangement leads to the low-bandwidth system behaviour. In parallel with the spring/cylinder is a twin-pipe gas shock absorber (damper). Hydraulic fluid is
supplied to each corner via a pump regulated to 200 bars, and an accumulator at each axle
provides pressure levelling during peak loads. High-pressure fluid is carried by two valve
blocks one at each axle. Each block contains two 3-way proportional valves one for each
corner which are used to maintain plunger pressure and a 2-way shut-off valve for each corner to lock the cylinder when the vehicle is stopped and to provide failsafe operation. Return
fluid is damped through a return accumulator.
Figure 11. Schematic diagram of the Mercedes Active Body Control actuator.
1054
The ABC system utilises multiple sensors for mechanical state measurements. Sensors
include longitudinal and lateral accelerometers, three vertical accelerometers (for bump/
pitch/roll detections), as well as a level sensor and a plunger position sensor at each wheel.
Additionally, several hydraulic pressure sensors are used for pressure control and monitoring.
The ABC system employs various algorithm modules for vehicle dynamics control
depending on the vehicle states. Functions provided include speed-based load levelling and
ride height adjustment, skyhook control of the vehicle body heave mode, roll control for roll
angle response to vehicle lateral acceleration, and pitch control for pitch angle response to
longitudinal acceleration. The algorithms are further controlled by driver input via instrument panel height and performance mode switches. ABC has continued to evolve since its
first introduction in 1998. For example, the wheel damping is now continuously adjustable
and the efficiency of the pump has been increased. In addition, computational processing
power is more than double that of the previous system.[127]
5.5.2.
A broad bandwidth system was developed at Ford Research Laboratory in the early
1990s and was demonstrated by fitting the suspension concept hardware onto a 1989 Ford
Thunderbird.[128] The concept hardware and software verified the previous theoretical study
in ride quality improvement for rough road and for going over a speed bump or crossing
a crowned road. The effort also identified the shortcomings of the implemented hardware
structure to facilitate next generation active suspension hardware design.
This concept car development provided first-hand experience and useful answers to many
practical concerns such as actual power consumption, secondary ride harshness in the range
above 5 Hz, and actuator noise. It also enabled the identification of specific next steps to
address these issues. For example, some of the key items that have been identified to be
addressed in the future include the parasitic losses with the parallel spring structure and the
need for further refinement in the secondary ride quality.
The Ford Thunderbird BBAS research vehicle (Figure 12) used four high-fidelity electrohydraulic servo actuators, one at each corner of the vehicle. While a typical narrow bandwidth
suspension is controlled through three-way proportional pressure control valves, the BBAS
1055
was controlled through four-way servo valves, which have much higher precision and speed
of response. In addition, the BBAS actuators were based on double-acting cylinders capable
of equally fast rebound and jounce strokes. The vehicle also had one central and four cornerunit microprocessors for fast signal/control processing; four actuator displacement sensors
and four load cells for internal (force) loop calculations; and four air springs one at each
corner placed in parallel with the BBAS actuators. The air springs support and self-centre
the vehicle sprung mass as a typical load levelling set would do and at the same time provide
lower sprung mass natural frequency for more comfortable basic ride, which is then appropriately dynamically modified through the BBAS actuators. The system incorporated 26 various
sensors, including accelerometers, pressure sensors, vehicle speed sensors, and others.
The BBAS control strategy was based on coordinated individual wheel control and consisted of two hierarchical levels.[43,128] The outer loop level operated at a 20 ms rate. It
calculated the desired corner forces for the four BBAS actuators, desired operating modes
(handling or ride dominated), and checked the overall system integrity. The ride-related calculations were based on quarter-car vehicle models aimed at emulating skyhook damping
at each corner, which is often very close to the optimal possible ride benefit.[20] Different
effective spring and damping rates were used depending on prevailing operating modes, that
is, ride or handling. Table 1 summarises typical parameters used.[128]
The calculations also included roll and pitch attitude control during turning, braking, accelerating, and the like; this was based on lateral and longitudinal acceleration signals and
steering sensors. The commanded or desired levels of corner forces were communicated
to corner computers, which operated at higher sampling rates of 1 ms to implement the
inner-loop force control based on feedback provided by load cell sensors. The corner units
distributed the force requests between different servo valves. For example, the conventional
(relative) suspension damping and the passive segments of skyhook force were provided by
bypass or cross-port two-way servo valves that provided damping only when called for in a
manner similar to semi-active damping control.
The BBAS R&D vehicle accomplished many of its key objectives, especially with respect
to primary ride (up to circa 5 Hz) and handling. The typical results in terms of NASA ride
metrics are summarised in [128] where a smaller number on the scale of 10 implies better
ride. From this reference it can be seen that, on average, the active Thunderbird provided
the best ride among a dozen different vehicles under consideration. The only areas where the
BBAS had somewhat worse performance were on smooth roads, where Jaguar XJ6 and Lexus
LS 400 (both well known for their smooth ride) had an advantage. This may be attributable
to still less refined NVH (higher frequency) levels and related secondary ride quality that are
some of the remaining challenges for BBAS concepts. Based on their summarised results,
it should also be noted that the BBAS Thunderbird significantly outperformed both the
baseline passive suspension Thunderbird and its narrow bandwidth (NBAS) counterparts as
exemplified by production Infiniti Q45a and Toyota Soarer.
Table 1. Typical parameters used in Ford BBAS.
Ride mode
Parameter
Skyhook damping
Units
Lb-s/in
N-s/m
Conventional damping Lb-s/in
N-s/m
Suspension spring
Lb/in
N/m
Manoeuvring mode
Front
Rear
Front
Rear
23.5
4115.3
3.0
525.4
15.0
2626.8
10.0
1751.2
3.0
525.4
15.0
2626.8
23.5
4115.3
8.0
1400.9
120.0
21,014
10.0
1751.2
4.0
700.5
120.0
21,014
1056
In the same paper, the handling characteristics of the active Thunderbird are compared with
its passive (production) version. The results of back-to-back tests show that the active vehicle
results in more agile handling with faster yaw response, less roll and pitch, and significantly
larger maximum lateral acceleration (0.895 vs. 0.81 g). Thus, in a direct comparison between
the two vehicles, the active car displayed significant improvements in both vehicle ride and
handling.
While achieving the full potential of a BBAS still remains an elusive goal, significant
progress has been achieved with the above early efforts. This holds particularly true for the
case of primary ride and handling. The challenging areas in need of further development
include secondary ride, NVH, and system power consumption, complexity, and cost. Possible
approaches here have been outlined in many of the previously mentioned references (e.g.
[7,11,88,107,119]).
In closing, it should also be mentioned that in addition to the above electro-hydraulic
implementation of the BBAS there was an additional BBAS effort focused on Electrical
Active Suspension (EAS), which was capable of power/energy regeneration.[129] The EAS
concept was also successfully implemented in a prototype vehicle with similar results and
challenges as for the electro-hydraulic counterpart.
5.6.
Road preview feature using road information for semi-active and active control
The concept of utilising the road profile information prior to road disturbances hitting the
wheel, investigated in [3135], has been partially applied to production vehicles. The early
attempt includes the Nissan vehicles in 1990 with the Super Sonic Suspension [130] feature.
More recently, the preview feature can be found in the 2013 Mercedes S-Class with the
Magic Body Control feature.[131,132] The early Super Sonic Suspension was applied to a
semi-active suspension system and the recent Magic Body Control is equipped with an active
suspension.
The early Nissan Super Sonic Suspension had only three levels of damping that are automatically selected to provide comfort and handling based on the road surface profile scanned
with ultrasonic sensors, as well as steering angle, vehicle speed, brake on/off status, and
measured accelerations.
The Mercedes S-Class detects road surface undulations in advance using a stereo camera
and adjusts the suspension to deal with them accordingly. The suspension system adjusts its
shock absorber damping in advance, making it stiffer or softer for each individual wheel, and
uses active hydraulic pistons to control the desired wheel loads.
The preview information for a three-dimensional road profile is obtained through a stereo
camera that monitors an area up to 15 m ahead during good visibility preferably in the day
light with suitable road surface structures and at speeds up to 130 kph. The stereo camera
has an update rate of about 60 ms. Each section of the road can be measured several times
at various angles during vehicle approach, and statistical algorithms are used to provide the
final estimate of road height profile. It has been demonstrated that, under good visibility, the
system performs well, providing superior ride comfort [133] especially when going through
speed bumps and excellent handling during dynamic manoeuvres.
6.
Conclusions
A 2 DOF quarter-car model, equipped with an actuator that generates force between the
car body/sprung mass and wheel/unsprung mass, has been widely studied in the literature
1057
and in the process facilitated many useful insights and new concepts (sky hook, ground
hook, invariant points to name a few). Thus, this simple model served well as a base for
the comparison study of various suspension concepts, controller designs, benefits, and limitations. The present survey first discusses the various performance index/cost functions that
are important for the design of a passenger vehicle. It then reviews controller design limitations through the application of optimal control theory to a quarter-car model with an
LQ-based cost function. Figures are generated to illustrate the design limitations and tradeoffs in RMS measures of various vehicle attributes and their associated frequency contents.
These plots should help suspension designers to visualise the ability of active suspensions to
behave differently depending on the current (and forthcoming) road excitations and vehicle
manoeuvre-induced loading (e.g. braking, accelerating, and turning).
Based on more than 40 years of theoretical and practical developments in the industry and
academia, it can be concluded that one of the main challenges for widespread usage of active
suspensions still lies in the area of actuator design and implementation. It is expected that
superior outcome will eventually result from a synergetic interplay between control software
intelligence and ingenious hardware design, all led by model-based system engineering and
further facilitated by relentless developments in electronics, microcontrollers and associated
sensors, and vehicle-to-vehicle and vehicle-to-infrastructure connectivity. Indeed, an important part of such an actuator development should include realisation of the desired actuator
force or, alternatively, displacement. As mentioned before, producing and tracking the desired
force (or displacement) is by no means a trivial task, especially when combined with practical constraints on cost, packaging, weight, and energy consumption. This important aspect
of advanced suspension design should be carefully evaluated through computer simulations,
bench testing, and eventually full vehicle on-road verification.
The above review of current state of the art reveals that the next advances in active
and semi-active suspension design will mainly come from two thrust areas. The first is the
increased efficiency in actuator design and implementation such as the usage of systematic
control software algorithm design combined with the previously mentioned innovative hardware design measures such as inclusion of possibly variable in-series/parallel compliances
and fast load levelling. The second is more comprehensive usage of preview information from
camera, Global Positioning System (GPS), and electronic horizon such as vehicle-to-vehicle
communication, vehicle-to-infrastructure communication, vehicle localisation and real-time
accessing of cloud information, and crowd sourcing leading to up-to-date road profile maps.
Acknowledgement
The authors would like to thank their colleagues Li Xu, Vladimir Ivanovic, and Michael Fodor for their valuable
input and feedback on this survey.
Disclosure statement
No potential conflict of interest was reported by the authors.
References
[1] Galip UA, Peng H, Melih C. Automotive control systems. New York (NY): Cambridge University Press;
2012.
[2] Cao D, Song X, Ahmadian M. Editors perspectives: road vehicle suspension design, dynamics, and control.
Veh Syst Dyn. 2011;49(12):328.
[3] Lu J, Hrovat D, Thomas P, Engleman J, Tseng HE, Filev D. Adaptive crash height adjustment using active
suspensions. US patent 13/736,397. 2013.
1058
[4] Tseng EH, Hrovat D, Lu J, McConnell M, Dehmel M, Seemann M, Owens FV. Suspension control system to
facilitate wheel motions during parking. US Patent 8,825,292. 2014 Sep 2.
[5] Smith MC, Walker GW. Performance limitations and constraints for active and passive suspension: a
mechanical multi-port approach. Veh Syst Dyn. 2000;33(3):137168.
[6] Hrovat D. A class of active LQG optimal actuators. Automatica. 1982;18(1):117119.
[7] Hrovat D. Survey of advanced suspension developments and related optimal control applications. Automatica.
1997;33(10):17811817.
[8] Karnopp D, Crosby M, Harwood R. Vibration control using semi-active force generators. J Manuf Sci Eng.
1974;96(2):619626.
[9] Wilson D, Sharp R, Hassan S. The application of linear optimal control theory to the design of active
automobile suspensions. Veh Syst Dyn. 1986;15(2):105118.
[10] Yue C, Butsuen T, Hedrick JK. Alternative control laws for automotive active suspensions. In American
Control Conference; 1988 June 1517; Atlanta, GA.
[11] Sharp RS, Peng H. Vehicle dynamics applications of optimal control theory. Veh Syst Dyn. 2011;49(7):1073
1111.
[12] Smith CC, McGehee DY, Healey AJ. The prediction of passenger riding comfort from acceleration data.
ASME J Dyn Syst Meas Control. 1978;100(1):3441.
[13] Hedrick JK, Ravera RK, Amderes JR. The effect of elevated guideway construction tolerances on vehicle ride
quality. ASME J Dyn Syst Meas Control. 1975;97(4):408416.
[14] Fearnsides JJ, Hedrick JK, Firouztash H. Specification of ride quality criteria for transportation systems: the
state of the art and a new approach. High Speed Ground Transp J. 1974;8(2):125132.
[15] Hoberock LL. A survey of longitudinal acceleration comfort studies in ground transportation vehicles. ASME
J Dyn Syst Meas Control. 1977;99(2):7684.
[16] Anonymous. ISO: Mechanical vibration and shock: evaluation of human exposure to whole-body vibration.
ISO 2631, International Standard Organization; 1997.
[17] Smith CC. On using the ISO standard to evaluate the ride quality of broad-band vibration spectra in
transportation vehicles. ASME J Dyn Syst Meas Control. 1976;98(4):440443.
[18] Stephens GD. Comparative vibration environments of transportation vehicles. In: Proceedings of the Design
Engineering Technical Conference. Passenger Vibration in Transportation Vehicles; 1977 Sept 2628;
Chicago, IL. p. 5972.
[19] Hrovat D. Applications of optimal control to advanced automotive suspension design. ASME J Dyn Syst
Meas Control. 1993;115(2B):328342.
[20] Hrovat D. Chapter 31: Active and semi-active suspension control. In: Mastinu G, Plchl M, editors. Road and
off-road vehicle system dynamics handbook. Boca Raton (FL): CRC Press, Taylor and Francis Group; 2014.
p. 11791226.
[21] Boyd S, El Ghaoui L, Feron E, Balakishnan V. Linear matrix inequalities in system and control theory.
Philadelphia (PA): SIAM; 1994.
[22] Fialho I, Balas GJ. Road adaptive active suspension design using linear parameter-varying gain-scheduling.
IEEE Trans Control Syst Technol. 2002;10(1):4354.
[23] Becker G, Packard A. Robust performance of linear parametrically varying systems using parametricallydependent linear feedback. Syst Control Lett. 1994;23(3):205215.
[24] Chen H, Guo K. An LMI approach to multiobjective RMS gain control for active suspensions. In: Proceedings
of the 2001 American Control Conference; 2001 June 2527; Arlington, VA. p. 26462651.
[25] Tran M, Hrovat D. Application of gain-scheduling to design of active suspensions. In: Proceedings of the
1993 IEEE Conference on Decision and Control. Vol. 2; 1993 Dec 1517; San Antonio, TX. p. 10301035.
[26] Hedrick JK, Butsuen T. Invariant properties of automotive suspensions. Proc Inst Mech Eng D: J Automob
Eng. 1990;204(1):2127.
[27] Hrovat D, Margolis D. Realistic road-track systems simulation using digital computers. In: Proceedings of the
Winter Simulation Conference; 1975 Dec; Sacramento, CA. p. 481489.
[28] Hrovat D. Optimal passive vehicle suspensions [Ph.D. thesis]. Davis (CA): University of California;
1979.
[29] Tseng HE, Hedrick JK. Semi-active control laws-optimal and sub-optimal. Veh Syst Dyn. 1994;23(1):545
569.
[30] Giorgetti N, Bemporad A, Tseng HE, Hrovat D. Hybrid model predictive control application towards optimal
semi-active suspension. Int J Control. 2006;79(5):521533.
[31] Tomizuka M. Optimum linear preview control with application to vehicle suspension revisited. ASME J
Dyn Syst Meas Control. 1976;98:309315.
[32] Louam N, Wilson DA, Sharp RS. Optimal control of a vehicle suspension incorporating the time delay
between front and rear wheel inputs. Veh Syst Dyn. 1988;17(6):317336.
[33] Louam N, Wilson DA, Sharp RS. Optimization and performance enhancement of active suspensions for
automobiles under preview of the road. Veh Syst Dyn. 1992;21(1):3963.
[34] Hrovat D. Optimal suspension performance for 2D vehicle models. J Sound Vib. 1991;146(1):
93110.
[35] Hac A. Optimal linear preview control of active vehicle suspension. Veh Syst Dyn. 1992;21:167195.
[36] Hac A, Youn I. Optimal semi-active suspension with preview based on a quarter car model. J Vib Acoust.
1992;114(1):8492.
1059
[37] Hac A, Youn I. Optimal design of active and semi-active suspensions including time delays and preview. J
Vib Acoust. 1993;115(4):498508.
[38] Van Auken RM. Active suspension control synthesis using reduced order models that account for the time
delay between road inputs. In: ASME 8th Biennial Conference on Engineering Systems Design and Analysis;
2006 July 47; Torino, Italy. p. 451462.
[39] Jankovic M, Kolmanovsky I. Developments in control of time-delay systems for automotive powertrain
applications. In: Balachandran B, Kalmr-Nagy T, Gilsinn DE, editors. Delay Differential Equations. Berlin:
Springer; 2009. p. 5592.
[40] Thompson AG, Pearce C, Millers E. Direct computation of the performance index for an optimally controlled
active suspension with preview applied to a half-car model. Veh Syst Dyn. 2001;35(2):121137.
[41] Pilbeam C, Sharp RS. On the preview control of limited bandwidth vehicle suspensions. Proc Inst Mech Eng
D: J Automob Eng. 1993;207(3):185193.
[42] Prokop G, Sharp RS. Performance enhancement of limited-bandwidth active automotive suspensions by road
preview. IEE Proc Control Theory Appl. 1995;142(2):140148.
[43] Goran MB, Bachrach BI, Smith RE. The design and development of a broad bandwidth active suspension
concept car. In: XXIV FISITA Congress. Total Vehicle Dynamics. Vol. 2, no. IMECHE C389/418; 1992.
p. 231252.
[44] Smith MC. Achievable dynamic response for automotive active suspension. Veh Syst Dyn. 1995;24(1):
133.
[45] Smith MC, Wang FC. Controller parameterization for disturbance response decoupling: application to vehicle
active suspension control. IEEE Trans Control Syst Technol. 2002;10(3):393407.
[46] Wang FC, Smith MC. Disturbance response decoupling and achievable performance with application to
vehicle active suspension. Int J Control. 2002;75(12):946953.
[47] Brezas P, Smith MC. Linear quadratic optimal and risk-sensitive control for vehicle active suspensions. IEEE
Trans Control Syst Technol. 2014;22(2):543556.
[48] Hrovat D, Hubbard M. Optimum vehicle suspensions minimizing RMS rattlespace, sprung-mass acceleration
and jerk. J Dyn Syst Meas Control. 1981;103(3):228236.
[49] Davis BR, Thompson AG. Optimal linear active suspensions with integral constraint. Veh Syst Dyn.
1988;17(6):357366.
[50] Youn I, Im J, Tomizuka M. Level and attitude control of the active suspension system with integral and
derivative action. Veh Syst Dyn. 2006;44(9):659674.
[51] Karnopp D. Force generation in semi-active suspensions using modulated dissipative elements. Veh Syst Dyn.
1987;16(56):333343.
[52] Lin J, Kanellakopoulos I. Nonlinear design of active suspensions. IEEE Control Syst. 1997;17(3):
4559.
[53] Chantranuwathana S, Peng H. Adaptive robust force control for vehicle active suspensions. Int J Adapt
Control Signal Process. 2004;18(2):83102.
[54] Gordon TJ, Marsh C, Wu QH. A learning automaton methodology for control system design in active
vehicle suspensions. In: International Conference on Control 94. Vol. 1; 1994 Mar 2124; Coventry, UK.
p. 326331.
[55] Howell MN, Frost GP, Gordon TJ, Wu QH. Continuous action reinforcement learning applied to vehicle
suspension control. Mechatronics. 1997;7(3):263276.
[56] Valasek M, Kortum W. Nonlinear control of semi-active road-friendly truck suspension. In: Proceedings of
AVEC. Vol. 98; 1998 Sep 1418; Nagoya, Japan. p. 275280.
[57] Gordon TJ, Best MC. Dynamic optimization of nonlinear semi-active suspension controllers. In: International
Conference on Control 94. Vol. 1; 1994 March 2124, Coventry, UK. p. 332337.
[58] Gordon TJ, Sharp RS. On improving the performance of automotive semi-active suspension systems through
road preview. J Sound Vib. 1998;217(1):163182.
[59] Wang J, Wilson DA. Mixed GL2/H2/GH2 control with pole placement and its application to vehicle
suspension system. Int J Control. 2001;74(13):13531369.
[60] Abdellahi E, Mehdi D, MSaad M. On the design of active suspension system by Hinf and mixed H2/Hinf:
an LMI approach. In: Proceedings of the 2000 American Control Conference; 2000 June 2830; Chicago, IL.
p. 40414045.
[61] Lu J, DePoyster J. Multi-objective optimal suspension control to achieve integrated ride and handling
performance. IEEE Trans Control Syst Technol. 2002;10(6):807821.
[62] De Jager AG. Comparison of two methods for the design of active suspension systems. Optim Control Appl
Methods. 1991;12(3):173188.
[63] Yamashita M, Fujimoti K, Uhlik C, Kawatani R, Kumura H. Hinf control of an automotive active suspension.
In: Proceedings of the 29th IEEE Conference on Decision and Control; 1990 Dec 57; Honolulu, HI. p. 2244
2250.
[64] Balas GJ, Doyle JC, Glover K, Packard A, Smith R. Mu-analysis and synthesis toolbox, Matlab users guide.
Boston (MA): MathWorks Inc; 1993.
[65] Doyle J, Packard A, Zhou K. Review of LFTs, LMIs, and Mu. In: Proceedings of the 30th IEEE Conference
on Decision and Control. Vol. 2; 1991 Dec 1113; Brighton, UK. p. 12271232.
[66] Gobbi M, Mastinu G. Analytical description and optimization of the dynamic behaviour of passively
suspended road vehicles. J Sound Vib. 2001;245(3):457481.
1060
[67] Chatillon M, Jezequel L, Coutant P, Baggio P. Hierarchical optimization of the design parameters of a vehicle
suspension system. Veh Syst Dyn. 2006;44(11):817839.
[68] Chen H, Sun P, Guo K. A multi-objective control design for active suspensions with hard constraints. Proc
Am Control Conf. 2003;5:43714376.
[69] Georgiou G, Verros G, Natsiavas S. Multi-objective optimization of quarter-car models with a passive or
semi-active suspension system 45. Veh Syst Dyn. 2007;45(1):7792.
[70] Crews J, Mattson M, Buckner G. Multi-objective control optimization for semi-active vehicle suspensions. J
Sound Vib. 2011;330(23):55025516.
[71] Yi K, Hedrick J. Active and semi-active heavy truck suspensions to reduce pavement damage. SAE Technical
Paper 892486; 1989.
[72] Cole DJ, Cebon D. Truck suspension design to minimize road damage. Proc Inst Mech Eng D: J Automob
Eng. 1996;210(2):95107.
[73] Kitching KJ, Cole DJ, Cebon D. Theoretical investigation into the use of controllable suspensions to minimize
road damage. Proc Inst Mech Eng D: J Automob Eng. 2000;214(1):1331.
[74] Cole DJ. Fundamental issues in suspension design for heavy road vehicles. Veh Syst Dyn. 2001;35(45):
319360.
[75] Novak M, Valasek M. A new concept of semi-active control of trucks suspension. In: Proceedings of
AVEC96; 1996 June 2428; Aachen, Germany. p. 141152.
[76] Valek M, Novak M, ika Z, Vaculin O. Extended ground-hook-new concept of semi-active control of trucks
suspension. Veh Syst Dyn. 1997;27(56):289303.
[77] Valek M, Kortm W, ika Z, Magdolen L, Vaculn O. Development of semi-active road-friendly truck
suspensions. Control Eng Pract. 1998;6(6):735744.
[78] Vaculn O, Valek M, Kortm W. Multi-objective control of semi-active truck suspensions. In: Proceedings
of AVEC2000; 2000 August 2224; Ann Arbor, MI. p. 405412.
[79] Valasek M, Sveda J, Sika Z. Soil-friendly off-road suspension. Veh Syst Dyn. 1998;44(Suppl 1):
479488.
[80] Valasek M, Kejval J, Maca J, Smilauer V. Bridge-friendly truck suspension. In: Proceedings of the 18th
IAVSD Symposium. Vol. 41; 2003 Aug 2529; Atsugi, Japan. p. 1322.
[81] MVO, Valek KJ. Global chassis control: integration synergy of brake and suspension control for active
safety. In: Proceedings of the International Symposium on Advanced Vehicle Control; 2004 Aug 2327;
Arnhem, The Netherlands. p. 495500.
[82] van Zanten A. Control of horizontal vehicle motion. In: Mastinu G, Plchl M, editors. Road and offroad vehicle system dynamics handbook. Boca Raton (FL): CRC Press, Taylor and Francis Group; 2014.
p. 10931178.
[83] Papageorgiou C, Smith MC. Positive real synthesis using matrix inequalities for mechanical networks:application to vehicle suspension. IEEE Trans Control Syst Technol. 2006;14(3):423435.
[84] Smith MC. Synthesis of mechanical networks: the inerter. IEEE Trans Automat Control. 2002;47(10):
16481662.
[85] Karnopp DC, Margolis DL, Rosenberg RC. System dynamics: a unified approach. 3rd/8th ed. New York:
Wiley; 2000 (3rd)/ 2012 (8th).
[86] Margolis D, Wilson R. Bond graph modeling of fluid filled engine mounts. Simulation Series 31: Proceedings
of the 1999 International Conference on Bond Graph Modeling and Simulation (ICBGM 99): [held as part
of] 1999 Western Multi Conference; 1999 January 1720; San Francisco, CA. p. 211216.
[87] Singh R, Adiguna H, Tiwari M, Tseng H, Hrovat D. Transient response evaluation of a hydraulic engine
mount. J Sound Vibration. 2002;36(7):3035.
[88] Evangelou S, Limebeer DJN, Sharp RS, Smith MC. Control of motorcycle steering instabilities. IEEE Control
Syst Mag. 2006;26(5):7888.
[89] Smith MC, Wang FC. Performance benefits in passive vehicle suspensions employing inerters. Veh Syst Dyn.
2004;42(4):235257.
[90] Gillespie TD. Fundamentals of vehicle dynamics. Warrendale (PA): Society of Automotive Engineers;
1992.
[91] Malek KM, Hedrick JK. Decoupled active suspension design for improved automotive ride quality/handling
performance. Veh Syst Dyn. 1986;15(Suppl 1):383398.
[92] Hayakawa K, Matsumoto K, Yamashita M, Suzuki Y, Fujimori K, Kimura H. Robust H-output feedback control of decoupled automobile active suspension systems. IEEE Trans Automat Control. 1999;44(2):
392396.
[93] Venhovens PJT. Optimal control of vehicle suspensions [Ph.D. Thesis]. Delft: Delft University of Technology;
1993.
[94] Best M, Gordon T, Dixon P. An extended adaptive Kalman filter for real-time state estimation of vehicle
handling. Veh Syst Dyn. 2000;34:5775.
[95] Jeppesen BP, Cebon D. Analytical redundancy techniques for fault detection in an active heavy vehicle
suspension. Veh Syst Dyn. 2004;42(12):7588.
[96] Nastasic , Dek Jahn G. Citroen technical guide [Internet]. Available from: http://www.citroenkerho.fi/xantia
/pdf/tekniikka/Tekniikkaopas.pdf
[97] Chance BK. 1984 Continental mark VII/Lincoln continental electronically-controlled air suspension (EAS)
System. SAE Technical Paper 840342; 1984.
1061
[98] Edmunds D. 2012 Tesla Model S signature performance suspension walkaround [Internet]. 2012 Sep 26 [cited
2015 Mar]. Available from: http://www.edmunds.com/car-reviews/track-tests/2012-tesla-model-s-signatureperformance-suspension-walkaround.html
[99] Korosec K. Potholes and Teslas model S: never the twain shall meet [Internet]. 2014 Sep 19 [cited 2015
Mar]. Available from: http://fortune.com/2014/09/19/tesla-model-s-suspension-upgrade/
[100] Kojima H, Nakano J, Nakayama H, Kawashima N, Fujimoto H. Development of new Toyota electronic
modulated suspension-two concepts for semi-active suspension control. SAE Technical Paper 911900;
1991.
[101] Nicolas R. Intelligent suspension system of Lincoln MKZ help mitigating pothole damage [Internet].
2014 Mar 20. Available from: http://www.car-engineer.com/intelligent-suspension-system-lincoln-mkz-helpmitigating-pothole-damage/
[102] Lincoln Drive Control: CCD. Lincoln Motor Company [Internet]. Available from: http://www.lincoln.com/
cars/mks/features/?featID = 1#page = Feature1
[103] Bodie MO, Hac A. Closed loop yaw control of vehicles using magneto-rheological dampers. SAE Technical
Paper 2000-01-0107; 2000.
[104] Hac A, Bodie MO. Improvements in vehicle handling through integrated control of chassis systems. Int J Veh
Autonomous Syst. 2002;1(1):83110.
[105] Fodor M, Redfield RC. Resistance control, semi-active damping performance. In: ASME Dynamic Systems
and Control Division, International Mechanical Engineering Congress and Exposition; 1995 Nov. 1217; San
Francisco, CA. p. 161169.
[106] Fodor M, Redfield RC. Experimental verification of resistance control, semi-active damping. Veh Syst Dyn.
1996;26(2):143159.
[107] Sharp RS. Variable geometry active suspension for cars. Comput Control Eng J. 1998;9(5):217222.
[108] Watanabe Y, Sharp RS. Mechanical and control design of a variable geometry active suspension system. Veh
Syst Dyn. 1999;32(23):217235.
[109] Tumova G. Delft active suspension (DAS-II) [Internet]. Research report OND1281145. Available from:
http://www.narcis.nl/research/RecordID/OND1281145
[110] Venhovens PT, van der Knaap AM. Delft active suspension (DAS) background theory and physical realization. In: Pauwelussen JP, Pacejka HB, editors. Smart vehicles. Amsterdam: Swets & Zeitlinger Publishers;
1995. p. 139165.
[111] van der Knapp A. Design of a low power anti-roll/pitch system for a passenger car. Delft University of
Technology, Vehicle Research Laboratory, Report 89.3VT.2628.
[112] Venhovens P, van der Knaap A, Pacejka H. Semiactive vibration and attitude control. In: Proceedings of the
International Symposium on Advanced Vehicle Control, AVEC 92; 1992 Sept 1417; Yokohama, Japan.
p. 170175.
[113] Evers WJ, Besselink IJM, van der Knaap ACM, Nijmeijer H. Analysis of a variable geometry active suspension. In: Proceeding of International Symposium on Advanced Vehicle Control; 2008 Oct 69; Kobe
International Conference Center, Kobe, Japan. p. 350355.
[114] Packer MB. Active ride control a logical step from static vehicle attitude control. SAE Technical Paper
780050; 1978.
[115] Crolla DA, Horton DNL, Pitcher RH, Lines JA. Active suspension control for an off-road vehicle. Proc Inst
Mech Eng D: J Automob Eng. 1987;201(1):110.
[116] Raad J, Oliver J, Norton R, Kokotovic V. Method of improving response time in a vehicle active tilt control
system. US Patent 5,948,028. 1999.
[117] Raad J, Oliver J, Rhode D, Norton R, Hermann S, Watson J. Method of controlling a vehicle active tilt control
system. US Patent 6,076,027. 2000.
[118] Konik D, Bartz R, Barnthol F, Bruns H, Wimmer M. Dynamic drive the new active roll stabilization system from BMW group system description and functional improvements. In: Proceeding of International
Symposium on Advanced Vehicle Control; 2000 Aug 2224; Ann Arbor, MI. p. 413420.
[119] Hrovat D, Tseng E, Fodor M, Asgari J. Chapter 22: Active and semiactive suspension systems. In: Mastinu
G, Plchl M, editors. Road and off-road vehicle system dynamics handbook. Boca Raton (FL): CRC Press,
Taylor and Francis Group; 2014. p. 769796.
[120] Hillebrecht P, Konik D, Pfeil D, Wallentowitz H, Zieglmeier F. The active suspension between customer
benefit and technological competition. In: Proceedings of XXlV FISITA Congress. Total Vehicle Dynamics.
Vol. 2; 1992 June 711; London. p. 221230.
[121] Akatsu Y, Fukushima N, Talcahashj K, Satch M, Kawaranki Y. An active suspension employing an
electrohydraulic pressure control system. SAE Technical Paper 905123; 1990.
[122] Alleyne A, Liu R. On the limitations of force tracking control of hydraulic servosystems. ASME J Dyn Syst
Meas Control. 1999;121(2):184190.
[123] Zhang Y, Alleyne A. A practical and effective approach to active suspension control. In: Proceedings
of the 6th International Symposium on Advanced Vehicle Control; 2002 Sep 913; Hiroshima, Japan.
p. 153158.
[124] Peng H, Strathearn R, Ulsoy A. A novel active suspension design technique simulation and experimental
results. In: Proceedings of the 1997 American Control Conference. Vol. 1; 1997 June 46; Albuquerque, NM.
p. 709713.
1062
[125] Merker T, Gaston G, Olaf T. Active Body Control (ABC) the DaimlerChrysler active suspension and damping
system. SAE Technical Paper 2002-21-0054; 2002.
[126] Merker T, Wirtz J, Hill M, Jeglitzka M. Das SL Fahrwerk-Dynamik und Komfort vereint (The SL chassisdynamics and comfort combined). ATZ/MTZ Magazine Special Issue; 2001. p. 8491.
[127] Kruse J. Mercedes-Benz NEXT online magazine [Internet]. 2014 June 4. Available from: http://next.mercedes
-benz.com/en/mbc-eng/
[128] Goran M, Smith R. Insights gained from active suspension development. In: Proceedings of the 26th FISITA
Congress; 1996 June 1623; Prague, Czech Republic. p. 24862514.
[129] Davis RI, Patil PB. Electrically powered active suspension for a vehicle. US Patent 5,060,959; 1991 Oct 29.
[130] Sugasawa F, Kobayashi H, Kakimoto T, Shiraishi Y, Tateishi Y. Electronically controlled shock absorber
system used as a road sensor which utilizes supersonic waves. SAE Technical Paper 851652; 1985.
[131] Streiter IR. Active preview suspension system. ATZ Worldwide. 2008;110(5):411.
[132] Rauh J, Ammon D. System dynamics of electrified vehicles: some facts, thoughts, and challenges. Veh Syst
Dyn. 2011;49(7):10051020.
[133] Mercedes-Benz S550. Consumer reports; 2014 Sep. p. 53.
To cite this article: Emir Kutluay & Hermann Winner (2014) Validation of vehicle dynamics
simulation models a review, Vehicle System Dynamics: International Journal of Vehicle Mechanics
and Mobility, 52:2, 186-200, DOI: 10.1080/00423114.2013.868500
To link to this article: http://dx.doi.org/10.1080/00423114.2013.868500
1.
Introduction
The simulation of vehicle dynamics has a wide array of applications in the development of
vehicle technologies, i.e. active suspensions, chassis design, controller design, driver assistance systems, development of simulators for ergonomics research, etc. Vehicle dynamics
simulations reduce the duration and costs during the research and development stages of new
designs and technologies.
Although simulation environments, measurement tools and mathematical theories on vehicle dynamics are well established, the methodical link between the experimental test data
and validity analysis of the simulation model is still lacking. This thesis aims to introduce a
methodology to be used in assessment of vehicle dynamics simulation models.
Validation of vehicle dynamics simulations is an intersection of two fields of study: simulation of vehicle dynamics and validation of simulations. Thus, the literature domain can be
divided into three main subjects: verification and validation (V&V) of computational models;
validation of vehicle dynamics simulation models in practice and validation methodologies
for vehicle dynamics simulation models.
In this paper, first the literature on V&V of computational models is presented. Definitions of
the V&V concepts, and approaches to the question from different disciplines are explored. Next
vehicle dynamics and the utilisation of simulation models in vehicle dynamics are explored.
The final part of the literature survey deals with the validation studies dedicated to simulation
of vehicle dynamics.
Corresponding
187
2.1.
Many definitions on V&V can be found in the literature. The most important concept and
definitions by various authors are presented in this subsection.
The goal of the V&V is to find out if a model is accurate when used to predict the performance
of the real-world system that it represents, or to predict the difference in performance between
two scenarios or two or more model configurations.[1] The process of verifying and validating
a model should also lead to improving a models credibility with users [2] and decisionmakers.[3]
Verification of a simulation model, in laymans terms, is defined as building the model
right. According to Carson,[1] verification is the techniques that are used to assure that
the model is correct and matches any agreed upon specifications and assumptions. Another
similar definition is the process of determining if a computational model of a physical event
and the code implementing the computational model can be used to represent the mathematical
model of the event with sufficient accuracy.[3] In the field of computational fluid dynamics
(CFD), the American Institute of Aeronautics and Astronautics (AIAA) definition is generally
accepted: The process of determining that a model implementation accurately represents the
developers conceptual description of the model and the solution to the model.[4] In another
source concerning CFD applications,[5] verification is defined as a process for assessing
simulation numerical uncertainty and estimating the sign, magnitude and uncertainty of the
simulation numerical error.
Same richness of definitions can also be encountered for the validation concept. Similar
to verification, in laymans terms, validation means Building the right model. One of the
earliest definitions is the process of confirming that the conceptual model is applicable or
useful by demonstrating an adequate correspondence between the computational results and
the actual other theoretical data.[6] Validation can be defined as the processes and techniques
that the model developer, model customer and decision-makers jointly use to assure that the
model represents the real system (or proposed real system) to a sufficient level of accuracy.[1]
Validation is the substantiation that a computerised model within its domain of applicability
possesses a satisfactory range of accuracy consistent with the intended application of the
model.[7] It is the process of determining if a mathematical model of a physical event represents
the actual physical event with sufficient accuracy.[3] According to AIAA,[4] validation is the
process of determining the degree to which a model is an accurate representation of the real
world from the perspective of the intended uses of the model.
Absolute validity is refuted by many experts.[2,3,8,9] A models validity is only defined
within the limits of the project and the intended application. Although a more comprehensive
validity analysis increases the credibility of the model, it also comes with extra financial and
time cost. Thus, a simulation model of a complex system can only be an approximation of the
actual system.[8] The logical conclusion is that, no matter how much time is spent to develop,
enhance and validate the model, there will always be discrepancies between the physical
phenomenon to be modelled and the simulation results.
2.2.
Philosophical aspect
The main question of concern is Can a simulation model be validated? when the philosophical perspectives to the simulation and validation are considered. This question is actually
very similar to one of the main problems of philosophy of science, regarding the scientific
progression.
188
189
mathematical model and then selecting the quantities of interest, statistical tolerances for
acceptance accordingly. On the second level, an iterative step takes place where the initial
findings are used to modify and enhance the model. According to this paradigm, verification is performed independent of the validation. Mathematical proof of convergence and the
reproducibility of the experimental results are the final two key concepts that are needed for
a healthy validation effort.
Another methodology for finite element simulation models is A-B-C-D Method, which
defines levels of V&V and approaches the problem from a cost-risk analysis aspect.[9] Here,
A stands for planning, B stands for solution verification, C stands for model validation and
D stands for model validation extrapolated out of the intended scope of application. This
approach introduces a scoring system for different levels of validation, acknowledging that
100% validation is impossible, and the level of attained validation is dependent on the scope
of the application. The needed level of validity comes with a cost to attain it, and this cost is
analysed depending on the application. Also, it has been noted during V&V analysis, that it
is better to use more than one method simultaneously instead of using one optimal method,
since every method has weaknesses and such a practice will remedy these and increase the
model credibility.[15]
In a study by Sarin et al.,[16] a methodology to construct a metric which is used to compare
time histories that are outputs of simulation models with time histories from experimental tests with emphasis on vehicle safety applications is established. The constructed metric
incorporates phase, magnitude and topology features in order to quantify the error between
the simulation and the experiment. Also a regression-based validation model was proposed
which uses this newly developed metric in order to reproduce the subjective judgements of
subject matter experts.
Romero worked on propagating system uncertainties into the simulation model through
model and data conditioning [17] and considers these as an essential step in model
validation.[18] The author considers the combined set of somewhat erred equations and associated compensating parameter values, and looks for effectiveness of the combined set, rather
than correctness of either or both. A model validation activity under representative conditions
is pursued to assess and to hopefully affirm the model, the conclusion being that in any real
validation experiment, there will be some uncertainty in the values of the actual inputs to the
system that is the subject of the model validation inquiry. The logic behind this conclusion
is that validation at the conditions of the validation experiment does not, in general, apply to
where the model will be used because of the different conditions of operation. Model validation and accuracy criteria are almost always substantially subjective and affiliated pass/fail
determinations are not sufficiently robust arbiters of model validity, quality and usefulness.
To extract the most value from validation experiments, any model bias and associated uncertainty should be accounted for in prediction. To accomplish this, a methodology to add the
uncertainty to the model in order to create an augmented or conditioned model that yields total
simulation uncertainty that is compatible with the uncertainty of the conditioned experimental
data was proposed.
Hypothesis testing and Bayesian statistical approaches are also researched as techniques to
validate simulation models. An enhanced Bayesian-based model validation method together
with probabilistic principal component analysis which uses Bayesian hypothesis testing and a
quantitative multivariate validation method based on probabilistic principal component analysis [19] and multivariate Bayesian hypothesis testing are proposed for simulation models
of dynamic systems.[20] These works focus on computer-aided engineering models of automotive safety applications (crash simulation and dummy passenger models), but have the
potential to find usage in vehicle dynamics as well. Using reversed hypothesis testing to validate methods [21] and employing statistical hypothesis testing as a form of objective cost-risk
190
analysis for validation of simulation models [22] are some examples to hypothesis testing
approaches to validation of simulation models. In these approaches the type II error (false
negative), which is the model users risk in modelling practice, is deemed more critical, since
accepting an invalid model as valid will result in the user making analysis with an invalid
simulation model and can lead to damages (even catastrophic results if, for example, the simulation model is for a construction project) and special emphasis is placed on minimising it.
Type I error (false positive) is the model builders risk, since rejecting a valid model will cost
extra work, time and money to the model building party, and does not have the potential to
cause any damage.
191
roll steer, deflection steer due to compliance and inaccurately calculates individual slip angles
of the tyres.[37] A survey on the modelling applications of the interaction between the tyre
tread-block and the road surface is provided by Wallaschek and Wies.[38]
This importance is further explored in another study with comparisons of tyre models with
different model depths.[39] The effects of different legal tyres on the same vehicle using
fishhook and sine-with-dwell manoeuvres are demonstrated by Arndt et al.[40] Up to 33%
discrepancy is observed for lateral acceleration gain between two approved tyres of the same
manufacturer.
One of the most importance sources on tyre dynamics is written by Pacejka,[41] who
also developed the so-called Magic Formula, an empirical tyre model which relies on curve
fitting using experimentally measured tyre data. Rill [42] developed a first order analytical
tyre model based on Taylor expansion of governing differential equations and another model
based on mechanical analogies is proposed by Lacombe.[43] Analytical models which use
modal parameters are also present.[44] Physical models, which use finite element model to
simulate the mechanics of the tyre structure, are very accurate but need substantial computing
power. Such models are not suitable for online usage. FTire is a recent example to this model
category.[45] Hybrid tyre models combine different model types to exploit their advantages.
A compromise between finite element models and analytical models which is based on the
macroscopic physical description of tyres is proposed by Gallrein and Bcker.[46] Similarly
analytical-empirical hybrid formulations are also existent.[47] Guo and Lu have developed a
semi-physical model, where a detailed nonlinear analytical description of tyre characteristics
is supported with experimentally measured tyre parameters with satisfactory results.[48] The
modelling of tyre wear is also an important aspect. Tyre wear is a major error source in
experimentation. Such models have uses in race performance prediction, tyre development
and fleet management.[49]
3.2. Practice of validation of simulation models for vehicle dynamics
Many of the publications which claim to present a validation methodology or technique tend
to only offer the application of a methodology to an individual case, like vertical dynamics of
articulated vehicles [50] or lateral dynamics of light vehicles.[51] These types of sources
are classified as project specific validation in vehicle dynamics and are explored in this
section together with other relevant research on the subject which do not present a validation
study.
Salaani et al. [52] and Heydinger et al. [53] worked intensively on development, parameter
measurement and validation of vehicle simulation models. A multibody full vehicle model
is developed, parameters for spring, damper, tyre and roll characteristics are measured and
curve fits are generated using these measurements. The performed evaluation covers vehicle
directional dynamics that include steady-state, transient and frequency domain responses. It
is concluded that, any detected discrepancy can be caused by a number of reasons including
model formulation, programming, parameter identification and experimental procedures; and
that the comparison analysis should be supported with analytical reasoning and common sense,
which is a subjective approach.
The methodology consists of three main phases: experimental field data collection, independent vehicle parameter measurement and model formulation, and comparison of simulation
predictions with field data using the same driver control inputs. The importance of independent parameter measurement is emphasised. The model parameters should not be adjusted
according to field tests to obtain a match. The comparisons are performed in time domain to
check the steady state and low-frequency responses and nonlinear effects; and in frequency
domain to check the high-frequency dynamics during transient manoeuvres.
192
The manoeuvres are so sequenced; first quasi-steady state, then step response, then pulse
response (evaluated in frequency domain) and finally a purpose-dependent real-world like
manoeuvre (lane change in this case) are performed. Confidence intervals are constructed,
but no validation criteria are defined using these confidence intervals. Furthermore, no validation metrics are constructed and the validation judgement is taken based on subjective
assessments with no quantitative foundations emphasising the adequacy of the simulation
model.
Validity analysis can also be employed for evaluation of identified vehicle parameters. The
application of genetic algorithm to the physical parameter estimation of a multibody vehicle
model for ride analysis is demonstrated in a project by Alasty and Ramezani [54] In this work,
the reference data are obtained using a more complicated multibody model. No metrics or
statistical analysis is utilised and the validation analysis is executed in time domain, although
the simulation model is developed for vehicle ride analysis. This conflict demonstrates the
importance of the planning and analysis of the simulation goals.
In another study by Mcnaull et al.,[55] a heavy truck simulation model is first modified
according to comparison of experimental and simulation results for lateral steady state manoeuvres; and then validated for dynamic response using a transient manoeuvre. The work does
not introduce or explain the methodology but rather is a demonstration that the end result
of the project is successful. Visual graphical comparison technique is used for validation,
but instead of overlaying the graphs, side-by-side placed diagrams are used, which diminish
the credibility of the validation judgement. Also, no metrics or statistical analysis is performed. The study demonstrates the correct way of using experimental data to correct the
simulation model, by determining the steady state offset and then testing the modified system with a transient manoeuvre. On the other hand applied validation technique, side-by-side
representation of quantities of interests, somewhat lowers the possibility of a healthy call for
validity.
Wade-Allen et al. discuss the validation of a full vehicle model in their 2002 paper.[56]
The research points out the importance of performing the parameter measurements in the
targeted operating regime. If the vehicle model is aimed for simulation of limit handling
scenarios, such as roll over or tyre saturation, the parameter measurements of the subsystems
of the simulation model must be accordingly measured, such as the tyre data over large slip
conditions and higher than normal load, and other nonlinearities due to larger deflections
caused by the highly dynamical manoeuvres.
The addressed validation issues include model formulation, verification of the computer coding, appropriate parameter estimation and measurement procedures and comparison between
experimental and simulation results. It is noted that a thorough validation analysis should
include both steady state and transient manoeuvres, evaluated in both time and frequency
domains. The model is tested using quasi-steady state steering wheel ramp input, pulse
response (in frequency domain), double lane change and fishhook manoeuvres. Validation
metrics or confidence intervals are not used and no statistical analysis is performed. A subjective and qualitative judgement is reached through visual graphical comparison of overlaid
time histories of test and simulation results.
This work reflects a correct approach to the validation problem, but with several shortcomings. The importance of parameter estimation and data validity is well emphasised, and
the sequencing of test manoeuvres, from steady state to transient and to real-life imitating
manoeuvres is proper. An alternative manoeuvre selection for frequency response can be sine
sweep, which will have the same power throughout the selected frequency range, contrary to
pulse response. However, transient response in time domain is not tested, and no quantitative criteria are set for validation. The validity judgement is taken according to the subjective
assessment of the visual resemblance of test and simulation results.
193
A similar study by Ozan et al. [57] is a typical example of the bountiful usage of the term
Validation Methodology. In this work a correlation methodology of a multibody simulation
model of a commercial vehicle is presented. A three-stage process is proposed. First, the vehicles suspension trimming at static conditions is implemented into the simulation model. The
second step is the extensive quasi-static testing of the kinematic components of the suspension and steering system, and their correlation to those of the simulation model. The last step
is the dynamical testing through linear swept steering manoeuvre and fishhook manoeuvre,
and visual graphical comparison of the time histories of the experiment measurements and
simulation outputs, without any statistical analysis, validation metric or accuracy criteria.
In summary, first the mass and properties, then kinematic modules of the simulation model
and finally the whole system response are checked using graphical representations. The
explained technique is project specific, and the used methodology to pass validation judgement lacks traceability and objectivity. The reasoning in the selection of manoeuvres used
in validation of the systems response is not explicit, and the assessment criteria are vague.
Neither validation metrics nor confidence intervals are constructed.
Another study by Hu [58] demonstrates the development of an analytical half vehicle suspension model for suspension control systems analysis and design. The model is validated
based on a comparison of an actual test vehicles and the models simulated time domain
responses over a particular road event which excites the low-frequency band of ride dynamics.
The model parameters are first fine-tuned using the results from a different experiment. This
practice is not advised in general and is only acceptable, if the data used in tuning and validation are different and independent.[53,59] In this study no validation metrics or statistical
analysis is performed. The results of the validation test are evaluated using visual graphical comparison by overlaying time histories of the experimental and model responses on the
same plot. However, the suspension, due its highly dynamic nature because of the constantly
changing vertical forces and the motion of the unsprung and sprung masses, is a subsystem
that should be analysed in the frequency domain. This work is a good example of an analysis
error and demonstrates why the analysis techniques and validity criteria should be defined and
documented at the start of the development project.
An approach to the validation problem as a multi-objective optimisation exercise is presented by Cassara et al.[60] Because of the large number of degrees of freedom and tuneable
parameters of the targeted simulation model, which is a Tractor-Semitrailer model for ride
and handling analysis, such an approach is proposed. To address this complicated question,
modal analysis is performed first at the component level and then at the subsystem level and
then the frequency response of the vehicle system is inspected. Manoeuvre odometrics are
checked as well, and several ride-related components are analysed in the frequency domain.
Focus of the research lies on the subsystem interaction and the effect of frequency response
modelling accuracy of subcomponents to the total system response considering ride and handling. No criteria are used to quantise the quality of the correlation between the experiments
and simulation results. Conclusions underline the importance of frequency domain agreement
of the subcomponents in isolating the problem zones in the simulation model.
A vehicle model/simulation evaluation tool for armed forces, called Model Post Processor is
developed by Howe et al.[61] The tool is capable of comparing different model structures with
each other or with actual static and dynamic test measurements for assessment and evaluation.
Evaluation of static metrics (mass properties, suspension kinematics, compliance, etc.) is
performed by a consistency check subroutine, and dynamic metrics are checked by another
subroutine through a range of test manoeuvres. Dynamics manoeuvre range encompasses
fundamental tests for longitudinal, lateral and vertical dynamics.
Horiuchi et al. have proposed a model-based validation procedure which evaluates the control and intervention performance for the certification of advanced chassis control systems.[62]
194
The proposed validation procedure handled the evaluation of the static and dynamic properties
in two consecutive steps. Braking stability is treated as a static property and is evaluated by
constrained bifurcation and continuation method, whereas the dynamical characteristics of
the vehicle are evaluated by the optimisation-based worst case scenario method. This work is
a good demonstration of application modelling and validation techniques to advanced chassis
control systems and virtual testing.
195
It is argued that due to the increasing complexity of modelling practices, it is generally not
possible to check all the equations (especially in multibody models) and numerical steps, and
running the simulation is the only way for verification.
This method proposes different validation approaches for different model depths. Closed
form solutions or estimates and the lateral load transfer measurements are compared
with the simulation results for manoeuvres lower than 0.5 g which do not involve brake
forces. This approach helps finding errors in inertial and geometric parameters, suspension stiffness concerning handling (cornering, aligning, steering, roll, etc.) and load transfer
model.
For higher than 0.5 g manoeuvres and manoeuvres with tyre saturation (limit handling),
checking the tyre forces as a function of kinematics and normal load is advised. This helps
detecting the errors in tyre model and suspension kinematics.
If the target application for the simulation model involves braking scenarios, checking
longitudinal load transfer, wheel slips, longitudinal tyre forces and, in the case of braking in
a turn manoeuvre, lateral tyre forces assist in finding the errors in longitudinal load transfer,
brake and tyre models.
This work criticises Heydinger et al. [59] for not only increasing confidence in the model,
but also accepting errors as long as the scatter is in an acceptable range, which would mask
the errors that stay within the defined interval. This view is supported with an example case:
an incorrect centre of gravity height measurement, naturally depending on the amount of
error, may provide sufficient results with respect to the confidence intervals for yaw rate and
lateral acceleration; but can be clearly detected by checking the lateral load transfer. On the
other hand, if a manoeuvre in which the lateral load transfer plays a significant role is the
target of the simulation project, such as fishhook manoeuvre, centre of gravity height, roll
angle and lateral load transfer states must be listed among the validation metrics at the start
of simulation project. This critic, therefore, should be directed to not the last stage of the
validation procedure, but to the planning stage, where the target manoeuvre is analysed and
test manoeuvres and validation metrics are chosen.
Another concern that could yield unreliable simulation results is the fact that the road friction
coefficient value supplied to the simulation is most of the time not the same value tested on the
actual test field. Determining or calibrating this value using data from the test vehicle taken
on the test field or directly implementing the manufacturer supplied values can lead to masked
errors.
Concerning the data validity, it is pointed out that faulty data entry is an important risk
factor, possible after the reliability of parameter measurements. According to Bernard and
Clover,[65] the most dangerous part in data entry of parameter values is tyre and suspension
data. Both the tyre model and the suspension model have many parameters, and this step is
prone to human error.
In a follow-up study by Gruening and Bernard,[66] data validity and faulty data entry problem is further investigated and some examples on the effects of different cases of faulty data
entry are demonstrated, although no general methodology to catch such errors is introduced. It
is suggested that unreasonable parameters may arise from three sources; erroneous measurements or bad guesses, misinterpretation of the parameters to be measured or mistakes in data
entry. Other than obvious recommendation of paying extra attention to data entry and checking for mistakes; a pre-processing procedure is suggested. Running the simulation through a
recipe of manoeuvres to determine metrics routinely associated with vehicle performance can
show some of the simple parameter errors, especially those associated with trim conditions
and steady state manoeuvres.
For example, vehicles generally have zero degrees of roll deflection at trim conditions. If a
simulation is run with straight driving at constant speed on a zero friction surface (thus, trim
196
condition), and roll angle is not zero, of course assuming that the mathematical equations of
the model are correct, that indicates that at least one parameter that affects static roll deflection
is wrong. However, even in this simple case, there would be more than one likely cause, for
example one of the parameters associated with tyre geometry or stiffness, or one of the spring
rates.
On the other hand, what if parameter data of one of spring rates and one of the tyres are
mistakenly entered at the same time, in such a way that their effects at trim condition cancel out
each other? In this case, a more dynamical manoeuvre (e.g. steady state cornering), individual
load or force measurements for the tyres, or an isolated test case would be more practical.
In the first two of these suggestions, there is absolutely no guarantee that the simultaneously
erroneously entered (or measured) parameters can be identified. Concerning the third suggestion, generally speaking, it is impossible to devise a manoeuvre which would isolate every
parameter of the system since most of the parameters are inherently interacting. One can only
come up with a limited number of such manoeuvres (for example lateral and longitudinal
manoeuvres can be separated, but the vertical dynamics almost always affect the other two)
but as previously said, there is no guarantee such an error can be detected. Nevertheless, this
approach is very useful in increasing the model confidence.
Allen et al. [67] provided a methodical approach to the validation problem. Possible problem areas causing inconsistencies between computer models and real world are
described as:
Mathematical model.
Computational model programming.
Parameter data.
Numerical accuracy and stability.
It is advocated that the vehicle dynamics model validation must be considered in context
and defined in terms of the domain of useful application, since a simulation model can only
be valid up to a degree and a model should be aimed for a certain behaviour, and a valid
model according to analysis of general system response does not guarantee valid subsystems
models.
The presented validation method is summarised in four steps:
For the validation in the lateral direction, three test cases, steady state cornering, sinusoidal
sweep and lane change manoeuvre, are chosen. The research does not offer a way to assess
the findings. Definition of validation metrics, application of statistical methods, or validity
criteria are not discussed.
Of these three approaches to validation methods for vehicle dynamics simulation models,
Heydinger et al. [59] and Garrott et al. [64] focus on operational validity and comparison of
test measurements and simulation results; Bernard and Clover [65] and Gruening and Bernard
[66] recommend analytical solutions and face validity checks for validation and vehicle tests
only for parameter identification and error hunt; and Allen et al. [67] emphasise importance
of face validity, analytical solutions and common sense checks with less methodical approach
to vehicle testing. A summary of these studies is presented in Table 1.
Reference
Handling
Handling
Ride
Handling
Handling
Handling
Hu [58]
Ride
Handling
Longitudinal,
handling and ride
Handling
Handling
4.
197
Method
Independent parameter measurement and real reference
data
Steady state, transient, frequency and real life like
manoeuvres
Analysis in time and frequency domains
Confidence intervals and no validation metrics
Independent parameter measurement and real reference
data
Steady state, step, frequency and real life like
manoeuvres
Analysis in time and frequency domains
Confidence intervals and no validation metrics
Reference data from a more complicated model
Analysis in time domain
No statistical analysis and no validation metrics
Model correction using experimental data
Transient manoeuvres
Analysis in time domain
No statistical analysis and no validation metrics
Parameter measurements at targeted operating regime
Steady state and transient manoeuvres
Analysis in time and frequency domains
No statistical analysis and no validation metrics
Independent parameter measurement and real reference
data
Transient manoeuvres
Analysis in time domain
No statistical analysis and no validation metrics
Model correction using experimental data
Analysis in time domain
No statistical analysis and no validation metrics
Modal analysis at subsystem level
Analysis in frequency domain
No statistical analysis and no validation metrics
Independent parameter measurement and real reference
data
Steady state, transient, frequency and real life like
manoeuvres
Analysis in time and frequency domains
Statistical analysis and validation metrics
Manoeuvre classification
Analysis in time and frequency domains
No statistical analysis and no validation metrics
Data validity and data entry errors
No manoeuvre selection
Analysis in time domain
No statistical analysis and no validation metrics
Data validity and data entry errors
Steady state, transient and real life like manoeuvres
Analysis in time domain
No statistical analysis and no validation metrics
Conclusion
198
Simulation and validation practices and methodologies in the field of vehicle dynamics are
presented.
There are many and similar definitions for V&V in the literature. One thing nearly all
experts agree upon is that an absolute validation is not possible, and validation analysis should
be handled according to the specific requirements and limitations of the application.
Existent works on validation methodologies for vehicle dynamics simulations focus on
different aspects of the question. Neither there is a standard in experimentation and data
handling processes in vehicle dynamics modelling, nor is there a standard reasoning process
in the vehicle dynamics modelling application in validation analysis. Most of the applications
rely only on visual comparison and subjective judgement. Diagrams types used in visual
comparison also do not follow any recognisable pattern and their contents and structures are
determined at will by the research team. Most of the time, the team which developed the model
also decides if the simulation is valid. This whole process chain diminishes the credibility of
these models.
A vehicle model should be analysed in time and frequency domain using both steady state
and transient manoeuvres. Both analyses can show characteristics which may go undetected
if only one is used. There are basic manoeuvres which demonstrate the general dynamics of
the vehicle, and then, there are other manoeuvres, which imitate real-life scenarios.
The validation criteria are dependent on the application; the validity metrics and the data
handling are dependent on the chosen manoeuvres and analysis; and the manoeuvre selection
is dependent on the targeted real-life phenomena to be simulated. Thus, validation of a vehicle
dynamics simulation model should be planned according to the characteristics of the real-life
phenomena it is trying to reproduce. The test manoeuvres to be used in the validation should
be determined early in the project timeline and according to the analysis of the real event to
be simulated.
To this aim a validation recipe, so to say, which incorporates a top-down approach is required.
Real-world manoeuvres and test manoeuvres need to be analysed, and the relationship in
between needs to be examined. For each test manoeuvre, experimental procedures, data handling techniques, validation metrics and accuracy criteria should be explicitly declared in
the project documentation. Such an approach will help to achieve more time and cost efficient simulation projects with increased model confidence by enhancing the traceability of the
validation process.
References
[1] Carson JS. Model verification and validation. Winter Simulation Conference; San Diego, CA; 2002.
[2] Sargent RG. Verification and validation of simulation models. Winter Simulation Conference; Austin, TX; 2010.
[3] Babuska I, Oden JT. Verification and validation in computational engineering and science: basic concepts.
Comput Meth Appl Mech Eng. 2004;193:40574066.
[4] American Institute of Aeronautics and Astronautics. Guide for the Verification and validation of computational
fluid dynamics simulations; AIAA; 1998. (G-077-1998e).
[5] Stern F, Wilson RV, Coleman HW, Paterson EG. Comprehensive approach to verification and validation of CFD
simulations part 1: methodology and procedure. J Fluids Eng. 2001;123:793801.
[6] Schlesinger S, Buyan JR, Callender ED, Clarkson WK, Perkins FM. Developing standard procedures for
simulation validation and verification. Summer Computer Simulation Conference; Houston, TX; 1974.
[7] Schlesinger S. Terminology for model credibility. Simulation. 1979;32(3):103104.
[8] Law AM. Simulation modeling & analysis. 4th ed. New York: McGraw-Hill; 2007.
[9] Logan RW, Nitta CK. Verification & validation: process and levels leading to qualitative or quantitative validation
statements. Warrendale (PA): SAE International; 2004. (SAE 2004-01-1752).
[10] Popper K. Logik der Forschung. 10th ed. Tbingen: Mohr Siebeck; 2005.
[11] Klein EE, Herskovitz PJ. Philosophical foundations of computer simulation validation. Simulat Gaming.
2005;36(3):303329.
[12] Tiller MM. Verification and validation of physical plant models. Warrendale (PA): SAE International; 2009.
(SAE 2009-01-0527).
199
[13] Blattnig SR, Green LL, Luckring JM, Morrison JH, Tripathi RK, Zang TA. Towards a credibility assessment of
models and simulations. 10th AIAA Non-Deterministic Approaches Conference; Schaumburg, IL; 2008.
[14] Oberkampf WL, Barone MF. Measures of agreement between computation and experiment: validation metrics.
J Comput Phys. 2006;217:536.
[15] Logan RW, Nitta CK. Comparing 10 methods for solution verification, and linking to model validation. J Aerosp
Comput Inf Commun. 2006;3:354373.
[16] Sarin H, Kokkolaras M, Hulbert G, Papalambros P, Barbat S, Yang RJ. A comprehensive metric for comparing
time histories in validation of simulation models with emphasis on vehicle safety applications. ASME International Design Engineering Technical Conferences and Computers and Information in Engineering Conference;
New York City, NY; 2008.
[17] Romero VJ. Type X and Y errors and data & model conditioning for systematic uncertainty in model calibration.
Warrendale (PA): SAE International; 2008. (SAE 2008-01-1368).
[18] Romero VJ. Validated model? Not so fast. The need for model conditioning as an essential addendum to model
validation. 48th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference;
Honolulu, HI; 2007.
[19] Fu Y, Zhan Z, Yang R. A study of model validation method for dynamic systems. Warrendale (PA): SAE
International; 2010. (SAE 2010-01-0419).
[20] Jiang X, Yang R, Barbat S, Weerappuli P. Bayesian probabilistic PCA approach for model validation of dynamic
systems. Warrendale (PA): SAE International; 2009. (SAE 2009-01-1404).
[21] Hartmann C, Smeyers-Verbeke J, Pennickx W, Heyden YV, Vankeerberghen P, Massart DL. Reappraisal of
hypothesis testing for method validation: detection of systematic error by comparing the means of two methods
or of two laboratories. Anal Chem. 1995;67(24):44914499.
[22] Balc S, Sargent RG. Some examples of simulation model validation using hypothesis testing. Winter Simulation
Conference; San Diego, CA; 1982.
[23] Ammon D. Modellbildung und Systementwicklung in der Fahrzeugdynamik. Stuttgart: B. G. Teubner; 1997.
[24] Diebold L, Schindler W, Haug J, Daesch C, Lahti M. Einspurmodell fr die Fahrdynamiksimulation und -analyse.
ATZ. 2006;11(108): 962967.
[25] Riekert P, Schunk TE. Zur Fahrmechanik des gummibereiften Kraftfahrzeugs. Ingenieur-Archiv. 1940;11:
210224.
[26] Olley M. Road manners of the modern car. Proc Inst Automobile Eng. 1946;41(1):523551.
[27] Segel L. Theoretical prediction and response of the automobile to steering control, research in automobile
stability and control in tire performance. Proc Automobile Div Inst Mech Eng. 1956;7:2646.
[28] McRuer D, Klein R. Automobile controllability driver/vehicle response for steering control volume 1
summary report. Warrendale (PA): SAE International; 1975. (NHTSA DOT HS-901 407).
[29] Schramm D, Hiller M, Bardini R. Modellbildung und simulation der dynamik von Kraftfahrzeugen. Berlin,
Heidelberg: Springer; 2010.
[30] Zomotor A. Fahrwerktechnik: Fahrverhalten. 2nd ed. Wrzburg: Vogel; 1991.
[31] Kiencke U, Nielsen L. Automotive control systems for engine, driveline and vehicle. 2nd ed. NewYork: Springer;
2007.
[32] Genta G. Motor vehicle dynamics. Singapore: World Scientific; 1997.
[33] Lugner P, Plchl M. Modelling in vehicle dynamics of automobiles. Z Angew Math Mech. 2004;84(4):219236.
[34] Blana E. A survey of driving research simulators around the world, institute of transport studies. Working Paper
481. University of Leeds; 1996 [cited 2013 Sep 12]. Available from: http://eprints.whiterose.ac.uk/2110/.
[35] Allen RW, Fancher PS Jr, Levison WH, Machey J, Mourant RR, Schnell T, Srinivasan R. Simulation and
measurement of driver and vehicle performance. Transportation in the New Millenium; 2000 [cited 2013 Sep
12]. Available from: http://trid.trb.org/view.aspx?id=639809.
[36] Allen RW, Rosenthal TJ. Requirements for vehicle dynamics simulation models. Warrendale (PA): SAE
International; 1994. (SAE 940175).
[37] Bundorf RT, Leffert RL. The cornering compliance concept for description of vehicle directional control
properties. Warrendale (PA): SAE International; 1976. (SAE 760713).
[38] Wallaschek J, Wies B. Tyre tread-block friction: modelling, simulation and experimental validation. Veh Syst
Dyn. 2013;51(7):10171026.
[39] Allen RW, Magdaleno RE, Rosenthal TJ, Klyde DH, Hogue JR. Tire modeling requirements for vehicle dynamics
simulation. Warrendale (PA): SAE International; 1995. (SAE 950312).
[40] Arndt MW, Rosenfield M, Arndt SM. How tires change a SUVs performance in fishhook and sine-with-dwell
testing. 21st International Technical Conference on the Enhanced Safety of the Vehicles; Stuttgart; 2009.
[41] Pacejka HB. Tyre and vehicle dynamics. 2nd ed. Oxford: Butterworth Heinemann; 2005.
[42] Rill G. First order tire dynamics. III European Conference on Computational Mechanics, Solids, Structures and
Coupled Problems in Engineering; Lisbon; 2006.
[43] Lacombe J. Tire model for simulations of vehicle motion on high and low friction road surfaces. Winter
Simulation Conference; Orlando, FL; 2000.
[44] Dihua G, Zhaolong D. Chengjian F. Tire model by using modal parameters directly (MPTM). Warrendale (PA):
SAE International; 2007. (SAE 2007-01-1512).
[45] Gipser M. FTire the tire simulation model for all applications related to vehicle dynamics. Veh Syst Dyn.
2007;45: 139151.
200
[46] Gallrein A, Bcker M. CDTire: a tire model for comfort and durability applications. Veh Syst Dyn.
2007;45(Supp.1):6977.
[47] Gim G, Choi Y, Kim S. A semi-physical tire model for a vehicle dynamics analysis of handling and braking.
Veh Syst Dyn. 2007;45(Supp.1):169190.
[48] Guo K, Lu D. UniTire: unified tire model for vehicle dynamic simulation. Veh Syst Dyn. 2007;45(Supp.1):7999.
[49] Braghin F, Cheli F, Melzi S, Resta F. Tyre wear model: validation and sensitivity analysis. Meccanica.
2006;41:143156.
[50] Cole DJ, Cebon D. Validation of an articulated vehicle simulation. Veh Syst Dyn. 1992;21(1):197223.
[51] Nalecz AG, Lu Z, DEntremont KL. Development and experimental validation of advanced dynamic vehicle
simulation (ADVS). Veh Syst Dyn. 1994;23(Supp.1): 390410.
[52] Salaani MK, Schwarz C, Heydinger GJ, Grygier PA. Parameter determination and vehicle dynamics modeling
for the national advanced driving simulator of the 2006 BMW 330i. Warrendale (PA): SAE International; 2007.
(SAE 2007-01-0818).
[53] Heydinger GJ, Schwarz C, Salaani MK, Grygier PA. Model validation of the 2006 BMW 330i for the national
advanced driving simulator. Warrendale (PA): SAE International; 2007. (SAE 2007-01-0817).
[54] Alasty A, Ramezani A. Genetic algorithm based parameter identification of a nonlinear full vehicle ride model.
Warrendale (PA): SAE International; 2002. (SAE 2002-01-1583).
[55] Mcnaull PJ, Guenther DA, Heydinger GJ, Grygier PA, Salanni MK. Validation and enhancement of a heavy
truck simulation model with an electronic stability control model. Warrendale (PA): SAE International; 2010.
(SAE 2010-01-0104).
[56] Wade-Allen R, Chrstos J, Howe G, Klyde DH, Rosenthal TJ. Validation of a non-linear vehicle dynamics
simulation for limit handling. Proc Inst Mech Eng D J Auto Eng. 2002;216:319327.
[57] Ozan B, sendur P, Uyank ME, z Y, Ylmaz SI. A model validation methodology for evaluating rollover
resistance performance of a ford commercial vehicle. Warrendale (PA): SAE International; 2010. (SAE-201001-0107).
[58] Hu H. Experimental validation of a half-vehicle suspension model. Warrendale (PA): SAE International; 1993.
(SAE 931966).
[59] Heydinger GJ, Garrott WR, Chrstos JP, Guenther DA. A methodology for validating vehicle dynamics
simulations. Warrendale (PA): SAE International; 1990. (SAE-900128).
[60] Cassara SJ, Anderson DC, Olofsson JM. A multi-level approach for the validation of a tractor-semitrailer ride
and handling model. Warrendale (PA): SAE International; 2004. (SAE 2004-04-2694).
[61] Howe JG, Chrstos JP, Romano R, OKins J. Development of a vehicle model/simulation evaluation tool.
Warrendale (PA): SAE International; 2008. (SAE 2008-01-0778).
[62] Horiuchi S, Okada K, Nohtomi S. Model-based validation procedure for the certification of advanced chassis
control systems. Veh Syst Dyn. 2010;48(Supp.1):393409.
[63] Hoskins AH, El-Gindhy M. Technical report: literature survey on driving simulator validation studies. Int J
Heavy Veh Syst. 2006;13(3):241252.
[64] Garrott WR, Grygier PA, Chrstos JP, Heydinger GJ, Salaani MK, Howe JG, Guenther DA. Methodology for
validating the national advanced driving simulators vehicle dynamics (NADSdyna). Warrendale (PA): SAE
International; 1997. (SAE 970562).
[65] Bernard JE, Clover CL. Validation of computer simulations of vehicle dynamics. Warrendale (PA): SAE
International; 1994. (SAE 940231).
[66] Gruening J, Bernard JE. Verification of vehicle parameters for use in computer simulation. Warrendale (PA):
SAE International; 1996. (SAE 960176).
[67] Allen RW, Rosenthal TJ, Klyde DH, Owens KJ, Szostak HT. Validation of ground vehicle computer simulations
developed for dynamics stability analysis. Warrendale (PA): SAE International; 1992. (SAE 920054).
Abstract
In recent years, driver drowsiness and distraction have
been important factors in a large number of accidents
because they reduce driver perception level and decision
making capability, which negatively affect the ability to
control the vehicle. One way to reduce these kinds of
accidents would be through monitoring driver and driving
behavior and alerting the driver when they are drowsy or in
a distracted state. In addition, if it were possible to predict
unsafe driving behavior in advance, this would also
contribute to safe driving. In this paper, we will discuss
various monitoring methods for driver and driving
behavior as well as for predicting unsafe driving behaviors.
In respect to measurement methods of driver drowsiness,
we discussed visual and non-visual features of driver
behavior, as well as driving performance behaviors related
to vehicle-based features. Visual feature measurements
such as eye related measurements, yawning detection,
facial expression are discussed in detail. As for non-visual
features, we explore various physiological signals and
possible drowsiness detection methods that use these
signals. As for vehicle-based features, we describe steering
wheel movement and the standard deviation of lateral
position. To detect driver distraction, we describe head
pose and gaze direction methods. To predict unsafe driving
behavior, we explain predicting methods based on facial
expressions and car dynamics. Finally, we discuss several
issues to be tackled for active driver safety systems. They
are 1) hybrid measures for drowsiness detection, 2) driving
context awareness for safe driving, 3) the necessity for
public data sets of simulated and real driving conditions.
1. Introduction
Recently, the total number of serious car crashes is still
increasing regardless of improvements in road and vehicle
design for driver safety. The U.S. National Highway
Traffic Safety Administration (NHTSA) data indicate that
more than 40,000 Americans suffer serious injuries from
56,000 sleep related road crashes annually [1]. According
to a study by the Sleep Research Center (UK), driver
drowsiness at the wheel causes up to 20% of accidents on
978-0-7695-5161-6/13 $31.00 2013 IEEE
978-1-4799-3022-7/13
DOI 10.1109/ICCVW.2013.85
616
617
Table 1: Summary of various features for detecting and predicting driver drowsiness.
618
619
5. Discussion
For active driver safety systems, we have discussed
various topics such as driver drowsiness detection, driver
distraction detection, and prediction of unsafe driving
behavior. To develop a better driver safety system, several
issues should be addressed. The most important ones are a)
hybrid measures for drowsiness detection, b) driving
context-awareness for safe driving, c) the necessity for
public data sets for simulation and real driving conditions.
620
6. Conclusion
Acknowledgements
References
[1] U.S. Dept. of Transportation, Traffic Safety Facts 2006: A
Compilation of Motor Vehicle Crash Data from the Fatality
Analysis Reporting System and the General Estimates
System, tech. report DOTHS 810 818, Natl Highway
Traffic Safety Administration, 2006.
[2] http://www.lboro.ac.uk/departments/ssehs/research/behavio
ural-medicine/sleep-research-centre/
[3] Y. Liang, Detecting driver distraction, Ph. D thesis,
University of Iowa, 2009.
621
622
623
Review Article
A Review of Active Yaw Control System for Vehicle Handling
and Stability Enhancement
M. K. Aripin,1 Yahaya Md Sam,2 Kumeresan A. Danapalasingam,2 Kemao Peng,3
N. Hamzah,4 and M. F. Ismail5
1
Control, Instrumentation & Automation Department, Faculty of Electrical Engineering, Universiti Teknikal Malaysia Melaka,
76100 Durian Tunggal, Melaka, Malaysia
2
Department of Control & Mechatronics, Faculty of Electrical Engineering, Universiti Teknologi Malaysia,
81310 UTM Johor Bahru, Johor, Malaysia
3
Temasek Laboratories, National University of Singapore 5A Engineering Drive 1, Singapore 117411
4
Faculty of Electrical Engineering, UiTM Pulau Pinang, 13500 Permatang Pauh, Pulau Pinang, Malaysia
5
Industrial Automation Section, Universiti Kuala Lumpur Malaysia France Institute, Section 14, Jalan Teras Jernang,
43650 Bandar Baru Bangi, Selangor, Malaysia
Correspondence should be addressed to M. K. Aripin; khairiaripin@utem.edu.my
Received 23 January 2014; Revised 8 May 2014; Accepted 15 May 2014; Published 12 June 2014
Academic Editor: Aboelmagd Noureldin
Copyright 2014 M. K. Aripin et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Yaw stability control system plays a significant role in vehicle lateral dynamics in order to improve the vehicle handling and stability
performances. However, not many researches have been focused on the transient performances improvement of vehicle yaw rate
and sideslip tracking control. This paper reviews the vital elements for control system design of an active yaw stability control
system; the vehicle dynamic models, control objectives, active chassis control, and control strategies with the focus on identifying
suitable criteria for improved transient performances. Each element is discussed and compared in terms of their underlying theory,
strengths, weaknesses, and applicability. Based on this, we conclude that the sliding mode control with nonlinear sliding surface
based on composite nonlinear feedback is a potential control strategy for improving the transient performances of yaw rate and
sideslip tracking control.
1. Introduction
In vehicle dynamic control of road-vehicle, controlling the
lateral dynamic motion is very important where it will
determine the stability of the vehicle. One of the prominent
approaches that are reported in the literature for lateral
dynamics control is a yaw stability control system. In order to
design an effective control system, it is essential to determine
an appropriate element of yaw stability control system. In
this paper, the elements of yaw stability control system, that
is, vehicle dynamic models, control objectives, active chassis
control, and its control strategies as depicted in Figure 1, are
extensively reviewed.
The linear and nonlinear vehicle models that described
the behaviour of lateral dynamic are explained for controller
design and evaluation purpose. To achieve the control objectives, it is essential to control the variables of yaw rate and
sideslip angle in order to ensure the vehicle stable. It is
required that the actual yaw rate and sideslip angle have
fast responses and good tracking capability in following
the desired responses. During critical driving condition or
manoeuvre, inappropriate commands by the driver to control
the steering and braking can cause the vehicle to become
unstable and lead to an accident. Therefore, an active control
for yaw stability control system is essential to assist the
driver to keep the vehicle stable on the desired path. By
implementing an active chassis control of steering or braking
or integration of both systems, the active yaw control system
can be realized.
Control objectives
Control strategies
Nonlinear model
7 DOF
8 DOF
Linearized model
14 DOF
2 DOF
d
Fx1 f
Fx2 f
Fy1
= (V + V )
= (1 + 2 ) sin + (1 + 2 ) cos + 3 + 4 .
(2)
Fy2
x
lf
r Mz
y
(3 + 4 ) + ,
Fx4
Fx3
(3)
lr
= V + V = V ( + ) .
Fy3
Fy4
(4)
1
[cos (cos (1 + 2 ) sin (1 + 2 ))
V
sin (sin (1 + 2 ) sin (1 + 2 ))]
.
(5)
1
[ ( cos + 2 cos + 1 sin + 2 sin )
1
(3 + 4 ) + ] .
(6)
(7)
Number of
Dynamic motions
DOF
7 DOF
(i) Longitudinal
(ii) Lateral
(iii) Vertical
(iv) Rotational of 4 wheels
8 DOF
(i) Longitudinal
(ii) Lateral
(iii) Vertical
(iv) Roll
(v) Rotational of 4 wheels
14 DOF
Fyr
Output variable
Fyf
f f
x
r
lf
lr
(i) Longitudinal
(ii) Lateral
(iii) Vertical
Yaw rate, roll rate, pitch
(iv) Roll
rate, and sideslip
(v) Pitch
(vi) Bounce
(vii) Rotational of 4 wheels
(viii) Vertical oscillations of
4 wheels
(8)
(9)
In this model, the front and rear lateral tire forces and ,
respectively, exhibit linear characteristics and described as a
linear function of the front and rear cornering stiffness,
and , as follows:
= ,
where the front and rear tire sideslip angle, , and for
linear tire forces are given in the following equations:
=
(iii) Left and right wheels at the front and rear axle are
lumped in single wheel at the centre line of the vehicle.
(iv) Constant vehicle speed i.e. the longitudinal acceleration equal to zero ( =0)
(v) Steering angle and sideslip angle are assumed small
( 0).
(vi) No braking is applied at all wheels.
(vii) Centre of gravity (CG) is not shifted as vehicle mass
is changing.
(viii) 2 front wheels have the same steering angle.
(ix) Desired vehicle sideslip is assumed to be zero in
steady state.
,
(11)
= + .
V
(10)
= ,
[ ] = [ 11 12 ] [ ] + [ 1 ] ,
21 22
2
1 +
]
[
V
V2
]
[
[ ]= [
][ ]
[
2 2 ]
[
(12)
[ V ]
]
[
+[
] ,
[ ]
[ ]
where and are state or output variables, and are front
and rear tire cornering stiffness, respectively, is vehicle
5
is essential to control both yaw rate and sideslip responses.
In order to achieve these control objectives, the proposed
controller must be able to perform the control task of the yaw
rate and sideslip tracking control.
Control objectives
Side slip
Yaw rate
V
( + ) + V2
(13)
where stability factor is depending on the vehicle parameters and defined as follows:
=
( )
( + )
(14)
[ ] = [ 11 12 ] [ ] + [ 1 ] + [ 3 ] ,
21 22
2
4
Active
braking
Active
differential
ARS
AFS
4WAS
fd
Desired
vehicle model
d , rd
Controller
Mz
, r
Actual vehicle
model
1 +
]
[
V
V2
]
[
[ ]= [
][ ]
2
2
[
]
[
0
[ V ]
]
[
+[
] + [ 1 ] .
[ ]
[ ]
[ ]
(15)
Although direct yaw moment control could enhance the
vehicle stability for critical driving conditions, it may be less
effective for emergency braking on split road surface. At
high vehicle speed steady state cornering, direct yaw moment
control could decrease the yaw rate and increase a burden
to the driver. To overcome this disadvantage, active steering
control is proposed.
4.2. Active Steering Control. Active steering control is another
approach to improving the vehicle yaw stability, especially
for steady state driving condition where the lateral tire
force is operated in the linear region. Research works of
active steering control have been continuously conducted in
order to improve the handling and stability performances as
reported in [7, 13, 3942]. In general, active steering control
can be divided into three categories, that is, active front
steering (AFS) control, active rear steering (ARS) control,
and four-wheel active steering (4WAS) control, as shown in
Figure 6. As road-vehicle normally has front-wheel steering,
AFS control becomes favourite approach among researchers
as it can be combined with active braking and/or suspension
control. In the AFS control diagram, as shown in Figure 8, the
[ ]=
1 +
]
[
V
V2
]
[
[
2
2 ] []
[
]
[
(16)
[ V ]
]
[
+[
] ( + ) .
[ ]
[ ]
On the other hand, ARS control is used to improve the
vehicle response for low speed cornering manoeuvres with
the input to the control system being the rear steering angle
. In order to enhance the manoeuvrability at low speed
and the handling stability at high speed, combination of AFS
control and ARS called 4WAS control has been proposed as
implemented in [24, 48, 49]. By implementing 4WAS control,
the lateral and yaw motion can be controlled simultaneously
using two independent control inputs. Noting that front
wheel steer angle and rear wheel steer angle with the
rear axles of rear tire cornering stiffness, and distance from
rear axle to centre of gravity are taken into account in the
input matric.
Desired
vehicle model
fd
e1 d
d , rd e = r r
2
d
fd + c
Controller
Actual vehicle
model
, r
fd
d , rd
e2
d
r rd
c
Controller
Mz
Actual vehicle
model
, r
[ ]=
[
V
[
[
[
[
[ V
[
+[
[
[
1 +
V2
2 2
V
]
]
] []
]
]
(17)
0]
[ V ]
]
]
[
] [ ] + [
] .
[ ]
1]
]
[ ]
Steering
Brakes
Vehicle actuator
Advantages
Disadvantages
(i) Less effective for braking on split road surface
(ii) Decrease yaw rate during steady state driving condition
(iii) Active differential need extra devices
8
International Journal of Vehicular Technology
9
theory (QFT) technique is implemented for robust active
front steering control in order to compensate for the yaw rate
response in presence of uncertainties parameters and reject
the disturbances [45]. As discussed in [48], robust controller
for 4WS vehicle is also designed based on -synthesis control
algorithm which considers the varying parameters induced
by the vehicle during driving conditions as uncertainties
while the study in [60] designed the steering control of vision
based autonomous vehicle based on the nested PID control
to ensure the robustness of the steering controller against the
speed variations and uncertainties of vehicle parameters.
In integrated active chassis control, an appropriate control
scheme is designed to meet the control objectives. Studies
in [2, 27, 33] have designed the control scheme that consists
of reference model based on linear parameter-varying (LPV)
formulation and static-state feedback controller with the
objective to ensure the robust performance for integrated
active front steering and active differential braking control.
In these studies, tire slip angle, longitudinal slips, and vehicle
forward speeds are represented as uncertainty parameters.
As reported in [4], integrated robust model matching chassis
controller that integrates active rear wheel steering control,
longitudinal force compensation, and active yaw moment
control is designed using controller based on linear
matrix inequalities (LMIs) for vehicle handling and lane
keeping performance improvement. In integrated active front
steering-direct yaw moment control, an optimal guaranteed
cost control (OGCC) technique is utilized in [10]. In this
study, tire cornering stiffness is treated as uncertainty during
variation of driving conditions. As discussed in [11], an
adaptive integrated control algorithm based on direct Lyapunov method is designed for integrated active front steering
and direct yaw moment control with cornering stiffness is
considered as a variation parameter to ensure the robustness
of designed controller. As reported in [23], sliding mode
controller is utilized for stabilising the forces and moments
in integrated control schemes that coordinated the steering,
braking, and stabiliser. In this study, the integrated control
structure is composed of a main loop controller and servo
loop controller that computes and distributes the stabilizing
forces/moments, respectively.
From the above discussion, these control strategies and
algorithms can be summarized and compared in terms of
their active chassis control, control objective, advantages,
and disadvantages as tabulated in Table 3. In conclusion,
an appropriate control strategy must be designed based
on particular algorithm. Robust control algorithms such as
, SMC, IMC, OGCC, QFT. are essential to solve the
uncertainties and disturbance problems that influenced the
yaw stability control performances. It is revealed that the
designed controllers in the above discussion are able to track
the desired yaw rate and vehicle sideslip response considering
external disturbances and system uncertainty.
Yaw rate
Yaw rate
Integrated AFS-DYC
Integrated AFS-DYC
DYC
4WAS
DYC
AFS
4WAS
OGCC
IMC
QFT
synthesis control
SMC-LQR
SMC-FLC
SMC
SMC-backstepping
Yaw rate
Yaw rate and sideslip
Yaw rate
sideslip
DYC
PID controller
Control objective
Control algorithms
Disadvantages
Transient response
improvement is not consider
Transient response
improvement is not consider
Advantages
Anti-wind-up strategy to avoid high
overshoot and large settling time
robust for uncertainties
10
International Journal of Vehicular Technology
1.4
1.2
Output response
system where the tire dynamic exhibit nonlinear characteristics, especially during critical driving conditions such as
a severe cornering manoeuvre. The main problems of yaw
rate and sideslip tracking control are uncertainties caused
from variations of dynamics parameters as discussed in the
previous section such as road surface adhesion coefficients
[8, 13, 33, 37, 45], tire cornering stiffness [2, 8, 1012, 20,
24, 30, 48, 49], vehicle mass [20, 30, 38, 45, 49], vehicle
speed [2, 13, 45], and moment of inertia [30, 49]. Besides
that, an external disturbance such as lateral crosswind may
influence the tracking control of desired yaw rate and
sideslip response as reported in [4, 6, 13, 24]. Therefore,
appropriate control strategies and algorithms are essential
to overcome these problems as discussed in the previous
section.
From the view of control system engineering, the
transient response performances of tracking control are
very important. However, the control strategies and algorithms discussed above are not accommodated for transient
response improvement of the yaw rate and sideslip tracking
control in presence of uncertainties and disturbances. The
designed controllers are only sufficient to track the desired
responses in the presence of such problems. Hence, an
appropriate control strategy that could improve the transient
performance of robust yaw rate and sideslip tracking control
should be designed for an active yaw control system which
can enhance the vehicle handling and stability performances.
11
1
0.8
0.6
0.4
0.2
0
10
12
14
16
18
20
Time (s)
Lightly damped system: fast rise-time and large overshoot.
Heavily damped system: sluggish response and small overshoot.
CNF control system: varying damping ratio,
fast and smooth response.
(18)
12
(, ) = () ,
(, ) = || ,
(, ) =
(19)
((1(0 )/(0 )) 1 ) .
1
1
1 ()
],
2 ()
(20)
where
1 := (, ) ,
(21)
8. Controller Evaluations
In order to evaluate the performance of designing controller,
simulations of emergency braking and driving manoeuvres
with the nonlinear vehicle model are usually carried out
according to ISO or SAE standards. The pure computer
simulations, cosimulation with other software or hardware in
the loop simulations (HILS), are the common approaches to
conducting the yaw stability test with or without driver model
for open loop or closed loop analysis, respectively.
One of the typical emergency braking manoeuvres for
vehicle yaw stability test is split- braking as reported in
[2, 37, 60]. In this test, the step input of brake torque is
applied to the vehicle in forward motion with constant speed
on split road surface adhesion coefficient, , where one side
of the wheels is on low and the other sides of the wheels
are on high or vice versa. This test is performed to test
the vehicle straight ahead driving stability. Critical driving
manoeuvres are also another efficient way to test the yaw
and lateral stability performances. A step steer manoeuvre
can be implemented to evaluate the steady state and transient
9. Conclusion
This paper has extensively reviewed the elements of yaw
stability control system. In designing yaw stability controller,
all these elements, that is, vehicle models, control objectives,
active chassis control, and control strategies, play an important role that contributes to the control system performances.
For controller design and evaluation, a 2 DOF linear and
7 DOF nonlinear vehicle models are essential. In order to
improve the handling and stability performances, the yaw rate
and sideslip tracking control are the main objectives that must
be achieved by the design controller. To realize an active yaw
stability control, an active chassis control of steering, braking
or integration of both chassis could be implemented with an
appropriate control strategies and algorithms.
In real driving condition, the uncertainties and external
disturbance may influenced the yaw rate and sideslip tracking
control performances. Hence, the robust control algorithm is
necessary. Based on this review, it has been concluded that
sliding mode control (SMC) is the best robust controller to
address these problems. From the view of control system,
transient performances are very important for tracking control. However, an existing SMC configuration does not have
capability to improve this transient performance. To address
this issue, a nonlinear sliding surface of SMC is designed
Conflict of Interests
The authors declare that there is no conflict of interests
regarding the publication of this paper.
Acknowledgments
The authors would like to thank to Ministry of Education of
Malaysia, UTeM, and UTM for the supports of the studies.
References
[1] B. Lacroix, Z. Liu, and P. Seers, A comparison of two control
methods for vehicle stability control by direct yaw moment,
Applied Mechanics and Materials, vol. 120, pp. 203217, 2012.
[2] S. C. Baslamisli, I. E. Kose, and G. Anlas, Handling stability
improvement through robust active front steering and active
differential control, Vehicle System Dynamics, vol. 49, no. 5, pp.
657683, 2011.
[3] H. Zhou and Z. Liu, Vehicle yaw stability-control system design
based on sliding mode and backstepping control approach,
IEEE Transactions on Vehicular Technology, vol. 59, no. 7, pp.
36743678, 2010.
[4] J. Wu, Q. Wang, X. Wei, and H. Tang, Studies on improving
vehicle handling and lane keeping performance of closedloop driver-vehicle system with integrated chassis control,
Mathematics and Computers in Simulation, vol. 80, no. 12, pp.
22972308, 2010.
[5] G. Tekin and Y. S. Unlusoy, Design and simulation of an integrated active yaw control system for road vehicles, International
Journal of Vehicle Design, vol. 52, no. 14, pp. 519, 2010.
[6] H. Ohara and T. Murakami, A stability control by active angle
control of front-wheel in a vehicle system, IEEE Transactions
on Industrial Electronics, vol. 55, no. 3, pp. 12771285, 2008.
[7] Y. Ikeda, Active steering control of vehicle by sliding mode
controlswitching function design using SDRE, in Proceedings
of the IEEE International Conference on Control Applications
(CCA 10), pp. 16601665, Yokohama, Japan, September 2010.
[8] H. Du, N. Zhang, and F. Naghdy, Velocity-dependent robust
control for improving vehicle lateral dynamics, Transportation
Research C: Emerging Technologies, vol. 19, no. 3, pp. 454468,
2011.
[9] B. L. Boada, M. J. L. Boada, and V. Daz, Fuzzy-logic applied
to yaw moment control for vehicle stability, Vehicle System
Dynamics, vol. 43, no. 10, pp. 753770, 2005.
[10] X. Yang, Z. Wang, and W. Peng, Coordinated control of AFS
and DYC for vehicle handling and stability based on optimal
guaranteed cost theory, Vehicle System Dynamics, vol. 47, no. 1,
pp. 5779, 2009.
[11] N. Ding and S. Taheri, An adaptive integrated algorithm for
active front steering and direct yaw moment control based on
direct Lyapunov method, Vehicle System Dynamics, vol. 48, no.
10, pp. 11931213, 2010.
13
[12] S.-B. Lu, Y.-N. Li, S.-B. Choi, L. Zheng, and M.-S. Seong,
Integrated control on MR vehicle suspension system associated
with braking and steering control, Vehicle System Dynamics,
vol. 49, no. 1-2, pp. 361380, 2011.
[13] S. Mammar and D. Koenig, Vehicle handling improvement by
active steering, Vehicle System Dynamics, vol. 38, no. 3, pp. 211
242, 2002.
[14] C. Zhao, W. Xiang, and P. Richardson, Vehicle lateral control
and yaw stability control through differential braking, in Proceedings of the International Symposium on Industrial Electronics
(ISIE 06), pp. 384389, July 2006.
[15] M. Mirzaei, A new strategy for minimum usage of external yaw
moment in vehicle dynamic control system, Transportation
Research C: Emerging Technologies, vol. 18, no. 2, pp. 213224,
2010.
[16] V. Cerone, M. Milanese, and D. Regruto, Yaw stability control
design through a mixed-sensitivity approach, IEEE Transactions on Control Systems Technology, vol. 17, no. 5, pp. 10961104,
2009.
[17] S. Zheng, H. Tang, Z. Han, and Y. Zhang, Controller design
for vehicle stability enhancement, Control Engineering Practice,
vol. 14, no. 12, pp. 14131421, 2006.
[18] E. Esmailzadeh, A. Goodarzi, and G. R. Vossoughi, Optimal
yaw moment control law for improved vehicle handling,
Mechatronics, vol. 13, no. 7, pp. 659675, 2003.
[19] M. Canale and L. Fagiano, Comparing rear wheel steering
and rear active differential approaches to vehicle yaw control,
Vehicle System Dynamics, vol. 48, no. 5, pp. 529546, 2010.
[20] M. Canale, L. Fagiano, A. Ferrara, and C. Vecchio, Comparing
internal model control and sliding-mode approaches for vehicle
yaw control, IEEE Transactions on Intelligent Transportation
Systems, vol. 10, no. 1, pp. 3141, 2009.
[21] S. Moon, W. Cho, and K. Yi, Intelligent vehicle safety control
strategy in various driving situations, Vehicle System Dynamics,
vol. 48, no. 1, pp. 537554, 2010.
[22] S. Yim, W. Cho, J. Yoon, and K. Yi, Optimum distribution of
yaw moment for unified chassis control with limitations on the
active front steering angle, International Journal of Automotive
Technology, vol. 11, no. 5, pp. 665672, 2010.
[23] D. Li, S. Du, and F. Yu, Integrated vehicle chassis control based
on direct yaw moment, active steering and active stabiliser,
Vehicle System Dynamics, vol. 46, no. 1, pp. 341351, 2008.
[24] T. Hiraoka, O. Nishihara, and H. Kumamoto, Automatic pathtracking controller of a four-wheel steering vehicle, Vehicle
System Dynamics, vol. 47, no. 10, pp. 12051227, 2009.
[25] S.-H. Yon, O.-S. Jo, S. Yoo, J.-O. Hahn, and K. I. Lee, Vehicle
lateral stability management using gain-scheduled robust control, Journal of Mechanical Science and Technology, vol. 20, no.
11, pp. 18981913, 2006.
[26] S. H. Tamaddoni, S. Taheri, and M. Ahmadian, Optimal
preview game theory approach to vehicle stability controller
design, Vehicle System Dynamics, vol. 49, no. 12, pp. 19671979,
2011.
[27] S. C. Baslamisli, I. E. Kose, and G. Anlas, Gain-scheduled
integrated active steering and differential control for vehicle
handling improvement, Vehicle System Dynamics, vol. 47, no.
1, pp. 99119, 2009.
[28] P. Falcone, H. Eric Tseng, F. Borrelli, J. Asgari, and D. Hrovat,
MPC-based yaw and lateral stabilisation via active front
steering and braking, Vehicle System Dynamics, vol. 46, no. 1,
pp. 611628, 2008.
14
[29] W. Cho, J. Yoon, J. Kim, J. Hur, and K. Yi, An investigation into
unified chassis control scheme for optimised vehicle stability
and manoeuvrability, Vehicle System Dynamics, vol. 46, no. 1,
pp. 87105, 2008.
[30] H. Du, N. Zhang, and G. Dong, Stabilizing vehicle lateral
dynamics with considerations of parameter uncertainties and
control saturation through robust yaw control, IEEE Transactions on Vehicular Technology, vol. 59, no. 5, pp. 25932597, 2010.
[31] Q. Li, G. Shi, J. Wei, and Y. Lin, Yaw stability control using the
fuzzy PID controller for active front steering, High Technology
Letters, vol. 16, no. 1, pp. 9498, 2010.
[32] W. J. Manning and D. A. Crolla, A review of yaw rate and
sideslip controllers for passenger vehicles, Transactions of the
Institute of Measurement and Control, vol. 29, no. 2, pp. 117135,
2007.
[33] S. C. Baslamisli, I. E. Kose, and G. Anlas, Design of active steering and intelligent braking systems for road vehicle handling
improvement: a robust control approach, in Proceedings of the
IEEE International Conference on Control Applications (CCA
06), pp. 909914, Munich 2006.
[34] P. Yih and J. C. Gerdes, Modification of vehicle handling
characteristics via steer-by-wire, IEEE Transactions on Control
Systems Technology, vol. 13, no. 6, pp. 965976, 2005.
[35] B. Kwak and Y. Park, Robust vehicle stability controller based
on multiple sliding mode control, in Proceedings of the SAE
World Congress, SAE 2001-01-10602001, 2001.
[36] P. Raksincharoensak, T. Mizushima, and M. Nagai, Direct yaw
moment control system based on driver behaviour recognition,
Vehicle System Dynamics, vol. 46, no. 1, pp. 911921, 2008.
[37] M. Canale, L. Fagiano, M. Milanese, and P. Borodani, Robust
vehicle yaw control using an active differential and IMC
techniques, Control Engineering Practice, vol. 15, no. 8, pp. 923
941, 2007.
[38] M. Canale, L. Fagiano, A. Ferrara, and C. Vecchio, Vehicle
yaw control via second-order sliding-mode technique, IEEE
Transactions on Industrial Electronics, vol. 55, no. 11, pp. 3908
3916, 2008.
[39] P. Falcone, F. Borrelli, J. Asgari, H. E. Tseng, and D. Hrovat,
Predictive active steering control for autonomous vehicle
systems, IEEE Transactions on Control Systems Technology, vol.
15, no. 3, pp. 566580, 2007.
[40] P. Falcone, F. Borrelli, H. E. Tseng, J. Asgari, and D. Hrovat, Linear time-varying model predictive control and its application to
active steering systems: stability analysis and experimental validation, International Journal of Robust and Nonlinear Control,
vol. 18, no. 8, pp. 862875, 2008.
[41] F. Borrelli, P. Falcone, T. Keviczky, J. Asgari, and D. Hrovat,
MPC-based approach to active steering for autonomous
vehicle systems, International Journal of Vehicle Autonomous
Systems, vol. 3, no. 24, pp. 265291, 2005.
[42] Y. Kawaguchi, H. Eguchi, T. Fukao, and K. Osuka, Passivitybased adaptive nonlinear control for active steering, in Proceedings of the 16th IEEE International Conference on Control
Applications (CCA 07), pp. 214219, October 2007.
[43] S. Singh, Design of front wheel active steering for improved
vehicle handling and stability, in Proceedings of the SAE
Automotive Dynamics & Stability Conference, SAE 2000-01-1619,
2000.
[44] W. A. H. Oraby, S. M. El-Demerdash, A. M. Selim, A. Faizz, and
D. A. Crolla, Improvement of vehicle lateral dynamics by active
front steering control, in Proceedings of the SAE Automotive
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
15
International Journal of
Rotating
Machinery
Engineering
Journal of
Volume 2014
The Scientific
World Journal
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
International Journal of
Distributed
Sensor Networks
Journal of
Sensors
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Volume 2014
Volume 2014
Journal of
Control Science
and Engineering
Advances in
Civil Engineering
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Volume 2014
Journal of
Robotics
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Volume 2014
VLSI Design
Advances in
OptoElectronics
International Journal of
Navigation and
Observation
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Chemical Engineering
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Volume 2014
Antennas and
Propagation
Hindawi Publishing Corporation
http://www.hindawi.com
Aerospace
Engineering
Volume 2014
Volume 2014
Volume 2014
International Journal of
International Journal of
International Journal of
Modelling &
Simulation
in Engineering
Volume 2014
Volume 2014
Volume 2014
Advances in
Volume 2014
ScienceDirect
Energy Procedia 82 (2015) 1017 1023
Abstract
Noise and vibration of automotive vehicles are an increasingly important issue in the automobile industry, for implications on both
environmental noise pollution and comfort perceived by driver and passengers. Since noise and vibration performances affect the
overall image of a vehicle, they are now considered important factors in the entire vehicle design process. In this regard, an accurate
experimental evaluation of vehicle noise and vibration levels in both stationary conditions and urban driving conditions are often
necessary for undertaking a refinement process of vehicle sound quality, satisfying the development targets. This paper provides a
review of the main experimental techniques adopted for the measurement and analysis of noise and vibration in a commercial
vehicle. In order to monitor sound generation in terms of both air-borne and structure-borne noise, the source, transfer path and
receiver have to be investigated by noise measurement. For this purpose, different techniques can be employed. Some of them make
use of instrumentation (i.e. condenser microphone, sound level meter, sound intensity probe, acoustic holography) which requires
particular measurement environments, as it ensures sufficient measuring accuracy only in free field or anechoic conditions. On the
contrary, the more innovative pressure-velocity sound intensity method can be used for in-situ tests, without the need to create
acoustically treated environments. At the same time, vibration measurements by using proper transducers allow to minimize the
structure-borne vibration transmitted from vehicle power plant to chassis and body structure. Experimental modal analysis is
another standard tool in vehicle NVH development for determining the dynamic characteristics of a system, and therefore for
reducing the risk of failure or excessively high structural vibration or sound pressure levels.
Hence, in present work, noise and vibration frequency spectra of experimentally analyzed case studies are given, to illustrate how
accurate experimental tests are fundamental for identifying the generation causes of typical noise and vibration problems.
2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
2015 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Selection and/or peer-review under responsibility of ATI
Peer-review under responsibility of the Scientific Committee of ATI 2015
Keywords:experimental tests, frequency analysis, noise, sound quality, vehicle refinement, vibration
1. Introduction
There are many sources of noise in a vehicle: in the past the engine was considered the most important, therefore
the first NVH studies were applied to reduce noise and vibrations generated by the engine and powertrain. As over the
years a strongly reduced level of noise has been reached for these systems, other sources of noise such as road noise
have become very significant. Besides, the increase of the vehicle speed has strongly enhanced the importance of
aerodynamic noise [1], as well. The engine provides a very important contribution to the annoying noise perceived
inside a vehicle, representing a great source of internal vibration, too. The vibrations derive from the reciprocating and
1876-6102 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Scientific Committee of ATI 2015
doi:10.1016/j.egypro.2015.11.861
1018
rotational masses such as pistons, connecting rods and shafts. Other sources of vibration come from the gearbox, the
differential and the structural vibrating modes of the exhaust system .In addition, there are many acoustic sources in
the powertrain system inducing airborne noise, and among them the most important are the intake and exhaust
tailpipes. They act as very efficient acoustic sources and can be considered as monopole sources. The vehicle
suspension system plays a key role in vibration transmission, being located in the structure-borne transmission path
between the roadtyre interaction and the vehicle body. The tyres have a dual role in road-noise generation and
transmission. On the other hand squeaks and rattles represent indirect noise sources, as they are generated by the
dynamic displacement of the surfaces of the dashboard and internal trimmings, caused by the overall vibrations in a
car. Besides the above main phenomena, there are other secondary noise sources, such as brakes, electrical and
mechanical accessories, etc.
It is important to consider that internal vehicle noise does not depend only on the acoustic and vibration sources;
key roles are also played by the different transmission paths between the sources and the receivers (i.e. the drivers and
passengers ears). In a vehicle there are two different categories of transmission paths, related to completely different
mechanisms of energy transmission: structure-borne and air-borne paths. Commonly in a car, the structure-borne noise
transmission path dominates at low frequency (<200 Hz) while the air-borne noise transmission path dominates above
500 Hz [2]. In the mid-frequency range, both transmission paths have usually the same level of importance.
In order to implement an improvement process of NVH performance of a vehicle, the knowledge of the main noise
sources and the transmission paths represents a fundamental aspect.
This paper provides an overview of the main experimental techniques employed for the identification of noise
sources in a vehicle. The description of the techniques will be supported by proper plots referred to experimentally
analyzed case studies under actual urban driving conditions. The aim is to point out the role of experimental activity in
identifying possible noise and vibration generation problems, which could affect vehicle sound quality and overall
harmony [3].
Nomenclature
AI
Articulation Index
NVH
EFR
EMA
f0
Ncyl
ODS
OMA
PVL
TPA
uref
tailpipe), structure-borne noise is caused from a vibrating source that induces the acoustic energy to travel through
solid structures and to be then released as air-borne noise (i.e. the engine structural vibrations). The main targets of
NVH experimental techniques are to determine the characteristics of these two types of noise in a vehicle, as well as
the transmission paths to driver and passengers. Only an accurate identification and characterization of noise and
vibration sources could allow to realize new refinements in design and to validate the solutions, in order to improve
vehicle interior sound comfort.
To date noise mapping techniques (sound intensity), acoustic holography and beam forming, Transfer Path
Analysis, modal analysis, order tracking are considered the most relevant experimental techniques to analyze and
identify NVH sources in a vehicle [4].Some of these techniques employ instrumentation which requires specifically
constructed rooms such as anechoic or reverberant chambers. In fact, only in these particular controlled sound fields,
the noisiness of a sound source can be easily related to the measured sound pressure ensuring sufficient accuracy. It is
the case, for instance, of a typical condenser microphone, that if used in rooms that are neither anechoic nor
reverberant could provide incorrect results, whether positioned too close to the sound source (at a distance less than the
wavelength of the lowest frequency emitted from the source, that is the near field) or too far away from the source,
where reflections from walls and other objects may considerably alter the measurement accuracy. Other measurement
devices that need to be employed in free field conditions are the sound level meter and the traditional pressure-pressure
sound intensity probe. However, thanks to the development of more innovative measurement techniques, sound
intensity in-situ tests are now possible also where anechoic conditions are not applicable, like an engine bay or car
interior. This technique relies on the use of the Microflown p-u intensity probe, the combination of a pressure
microphone and a particle velocity transducer which directly measure close to the source surface the scalar value
'sound pressure' and the vector value 'acoustic particle velocity' in one spot, giving in output the sound intensity
[5].Due to the near field benefits [6]using the direct measure of the directional particle velocity, p-u probes are not
highly affected by background noise or reflections, and measurements can be undertaken in real operational situation.
P-u probe represents an effective tool for vehicle noise sources identification issues. Typical measurements can be
performed, for instance, on the engine compartment adopting the Scan & Paint method [7]. The acoustic signals of the
sound field are acquired by manually scanning the emitting surface with the p-u probe while a camera is positioned
toward the surface to film the scanning. The sensor position is extracted by applying automatic color tracking to each
frame of the recorded video. It is then possible to directly visualize sound variations across the space in terms of sound
pressure, particle velocity or acoustic intensity.
As an example, in the following figures results obtained scanning the engine bay of a four-cylinder four-stroke
engine at a particular stationary operating condition (2000 rpm), are shown [8]. Fig. 1, in particular, reports acquired
data processed in terms of Particle Velocity Level (PVL), which is by definition equal to:
u
(1)
>dB@
PVL 20 log10
u ref
where u is the particle velocity magnitude and uref represents the reference effective particle velocity of 50 nano-meter
per second. The engine radiated noise spectrum always contains strong excitations associated with the cylinder firings
rate. The strongest tone in the spectrum is generally the Engine Firing Rate (EFR), defined as:
N
(2)
N cyl >Hz @
EFR
60 H
where N is the engine rotational speed expressed in revolutions per minute (rpm), is an integer equal to 1 or 2 for a
two-stroke or a four-stroke engine respectively, and Ncyl represents the number of cylinders. The engine firing
frequencies can be also expressed in terms of k-multiples of the rotating frequency, the so-called engine orders:
N
(3)
f o k >Hz @
60
According to formula (3) in a four-cylinder four-stroke engine the EFR is also called the 2ndengine order, as the
frequency is two times (k=2) the engines rotation one.
As expected, in Fig. 1 the first PVL peak which is the dominant excitation in the spectrum represents the 2nd engine
order, whereas the subsequent three peaks correspond to multiples of the engine combustion frequency at 2000 rpm.
1019
1020
This analysis can be also useful to identify occurrences not directly related to engine firings. For instance, in Fig. 1
the not negligible PVL peak around 500 Hzis likely due to a resonance effect of the system at this particular
frequency.
As a final result, it is possible to visualize the sound field across the scanned space. In this regard, Fig. 2 shows the
acoustic color map of particle velocity at the frequency of 66.4 Hz (the 2nd engine order). It is immediate to note that
the main acoustic emissions come from the internal combustion engine, which radiates up to about 95 dB.
This kind of sensor can be successfully used also for other applications in the field of mapping and analyzing
automotive noise and vibration. Panel Noise Contribution Analysis, a well-known methodology for an air-borne
Transfer Path Analysis (TPA) in car interior through which pressure contribution from the individual panels at a
reference point can be very accurately calculated [9][10], or Near Field Acoustic Camera for sound source localization
during transients even used in non-anechoic conditions [6], or In-Situ Absorption method for the measurement of the
acoustic properties of materials [11][12], are typical tools employed for increasing the acoustic comfort quality in a
car.
Order tracking analysis is another experimental technique which can provide useful information about vibrations
and acoustic signals generated in vehicle drive train components, related to engine rotational speed and in particular to
engine ignition frequency. In this respect, Fig. 3 shows an example of the spectrogram of exhaust noise signal acquired
by a proper microphone located in the proximity of the tailpipe end, during engine run-up from approximately 1000 to
3500 rpm [13].
Fig. 3. 3D acoustic intensity map of exhaust noise signal in engine run-up condition from approximately 1000 to 3500 rpm
Such analysis allows to visualize the magnitude peaks of the spectrum lined up with the orders spectrum lines, and
gives the possibility to identify the operating conditions and the frequency range of interest where noise emissions are
predominant. By observing Fig. 3, it can be noted that most of the noise content is produced in a low frequency range
up to approximately 350 Hz, with more pronounced peak values up to 100 Hz for engine rotational speeds above 2700
rpm. These latter could depend on a resonance effect related to exhaust natural frequencies to be better investigated.
Structural measurement techniques play a key role in the vehicle development process, as well. In particular,
powertrain vibration isolation design, if carefully carried out, allows to minimize the structure-borne vibration
transmitted from power plant to chassis and body structure. The most common acceleration transducer is the
piezoelectric accelerometer which consists of a quartz crystal with a mass bolted on top. Accelerometers are able to
detect very small vibrations and not be damaged by large vibrations in a wide frequency range from 0.2 Hz to 10 kHz.
The main methods used for structural dynamics analyses are Experimental Modal Analysis (EMA), Operational Modal
Analysis (OMA) and Operational Deflection Shapes (ODS).While the objective of traditional EMA is the construction
of a mathematical model of the dynamic characteristics of the object under investigation by measuring the frequency
response function between a force and a response transducers [14], OMA utilizes only response measurements of the
structure in operational condition to identify modal characteristics. On the other hand, ODS allows to determine the
real-world forced deflection during operation. With advances in performance of the test equipment and computer
capabilities, modal analysis has become a standard tool in vehicle NVH development. The knowledge of the relevant
dynamic characteristics helps engineers to avoid resonances at frequencies with high excitation, and therefore to
reduce the risk of damage or strong structural vibration or sound pressure levels.
3. Vehicle sound quality
As explained, the knowledge and characterization of the different noise sources play a key role in the sound
package design of a vehicle. However, although the level of interior noise reduction is very important for the acoustic
comfort of the occupants, it is not sufficient by itself. Another important factor that has to be taken into account is the
quality of the noise. In recent years there have been many sound quality studies and analyses in order to define the
criteria for a noise that is not just quiet but also pleasant. Some of the most commonly used sound quality and
psychoacoustic metrics for users perception in a vehicle are Articulation Index (AI), Loudness and Sharpness [2].
The AI is a measure of the intelligibility of voice signals, expressed as a percentage of speech units that are
understood by the listener when heard out of context. There are different algorithms for the calculation of the AI. The
method used in the automotive field is based on the 1/3-octave band levels in dB(A) between 200 Hz and 6300 Hz,
taking strongly into account the frequency bands where the typical speech frequencies are more important.
1021
1022
The Loudness level refers to the perceived intensity of the sound. It is based on the statistical response of the ear as
a function of frequency for a statistically significant number of people. The unit of Loudness is sone, which allows
to visualize the parameter on a linear scale. A sine tone of the frequency 1 kHz with a level of 40 dB has by definition
a Loudness of 1 sone. The Loudness scale is distinguished by the fact that a tone which is perceived to have double the
Loudness on the Loudness scale is designated by a doubled sone-value. For steady state sounds, standardized
calculation procedures have been defined by Zwicker and Stevens and accepted as ISO standards [15][16][17]. For
temporally varying sounds, Zwicker has also proposed an approach taking into account temporal effects [18]. The
latter evolved until its acceptance as DIN 45631/A1 [19].
The Sharpness is another important measure that takes into account the frequency content of the noise. It is defined
as the ratio between the high frequency noise level and the overall level; it is not related to sound intensity, but it is
high for metallic noise components that are generally considered annoying. A high Sharpness is hence correlated with
bad quality of the vehicle. The unit of Sharpness is acum. Sharpness delineates human sensation in a linear manner
as well. The reference sound of 1 acum is a narrow-band noise, one critical band wide, and at a center frequency of 1
kHz and having a level of 60 dB. The Sharpness calculation is based upon the specific Loudness distribution of the
sound. Therefore the calculation of the Loudness affects the results of the Sharpness analysis. The standard DIN 45692
standardizes the metric calculation.
An example of poor sound quality is represented by data reported in the following figures, [20]. In Fig. 4(a) the two
lines on the graph represent the Time Varying Loudness measured in a four-cylinder engine vehicle by two
conventional microphones located at the driver ears position, during engine run-up on a chassis dynamometer (red and
green indicate left and right ear respectively).As highlighted in the circled region, the overall level, measured in this
case in terms of Time Varying Loudness, exhibits strong deviations from the ideal trend (dashed blue line) at particular
engine speed ranges, giving the impression of a vehicle not characterized by a smooth and refined ride. In effect, a
vehicle signature that is loud but grows linearly with engine and vehicle speed will be more acceptable than a quieter
signature that shows large deviations from its mean trend under the same test conditions.
Fig. 4. Time Varying Loudness vs. RPM (a) and Sharpness vs. RPM (b) during vehicle acceleration
Also Sharpness plot in Fig. 4(b), which refers to the same test, indicates a low vehicle interior sound quality at a
particular operating condition (around 2675 rpm), as it reaches a substantially high level.
The measure of the mentioned psychoacoustic parameters helps to identify the best solutions to improve NVH
performance in terms of good interior sound quality rather than only reduced noisiness. Solutions can regard the use of
appropriate materials which ensure adequate acoustic insulation and absorption of the vehicle cabin, as well as
structural damping.
4. Conclusion
Several are the experimental techniques available nowadays for the characterization of vehicles noise and vibration.
In this paper a review of the main relevant experimental methods has been provided together with experimental data
representations of real analyzed case studies, with the aim to demonstrate how accurate measurements are very
important starting points for identifying the causes of noise generation and for undertaking a successive improvement
process of vehicles NVH performances.
References
[1] Yinzhi, H., Yang, Z., & Wang, Y. (2013). Wind Noise Testing at Shanghai Automotive Wind Tunnel Center. In Proceedings of the FISITA
2012 World Automotive Congress (pp. 571-578). Springer Berlin Heidelberg.
[2] Xu Wang, Vehicle noise and vibration refinement, Woodhead Publishing, 2010.
[3] G. Cerrato, Automotive Sound Quality Powertrain, Road and Wind Noise, Sound &Vibration, April 2009.
[4] G. Cerrato and P. Goodes, Practical approaches to solving noise and vibration problems, Sound & Vibration, April 2011.
[5] Microflown Technologies Ebook, Intensity.
[6] Microflown Technologies Ebook, Near Field Acoustics.
[7] Daniel Fernandez Comesana, Scan-Based Sound Visualisation Methods using Sound Pressure and Particle Velocity, Thesis for the degree of
Doctor of Philosophy, University of Southampton Institute of Sound and Vibration Research, July 2014.
[8] D. Siano, M. Viscardi, M. A. Panza, Experimental Acoustic Measurements in Far Field and Near Field Conditions: Characterization of a
Beauty Engine Cover, Recent Advances in Fluid Mechanics and Thermal Engineering, ISBN: 978-1-61804-268-2, pp.50-57, 2014.
[9] Bertolini, C., Horak, J., Mantovani, M., & Lo Sinno, G. (2011, July). Interior panel contribution based on pressure-velocity mapping and
acoustic transfer functions combined with the simulation of the sound package. In INTER-NOISE and NOISE-CON Congress and Conference
Proceedings (Vol. 2011, No. 6, pp. 2155-2172). Institute of Noise Control Engineering.
[10] Grosso, A., Fernandez Comesana, D. and De Bree, H., "Further Development of the PNCA: New Panel Noise Contribution ReferenceRelated (PNCAR)," SAE Int. J. Passeng. Cars - Mech. Syst. 5(2):2012, doi:10.4271/2012-01-1539.
[11] Li, M., van Keulen, W., Tijs, E., van de Ven, M., & Molenaar, A. (2015). Sound absorption measurement of road surface with in situ
technology. Applied Acoustics, 88, 12-21.
[12] Tijs, E., & Druyvesteyn, E. (2012). An intensity method for measuring absorption properties in situ. Acta Acustica united with Acustica,
98(2), 342-353.
[13] D. Siano, M. A. Panza, An Experimental Analysis of Vehicle Exhaust Noise Signals by Means of Order Tracking Approach, Recent
Advances in Systems, Proceedings of the 19th International Conference on Systems (part of CSCC '15), Zakynthos Island, Greece, July 16-20, 2015,
Recent Advances in Electrical Engineering Series, Vol.52, p.502, ISBN: 978-1-61804-321-4, ISSN: 1790-5117.
[14] D. Siano, M. Viscardi, P. Napolitano, M. A. Panza, Numerical and Experimental Acoustic Performance Investigations of a High-Speed
Train Composite Sandwich Panel, WSEAS TRANSACTIONS on APPLIED and THEORETICAL MECHANICS, E-ISSN: 2224-3429, Volume 9, pp.
290-300, 2014.
[15] International Organization for Standardization, Method for Calculating Loudness Level, ISO-532-1975 (E).
[16] E. Zwicker et al, Program for Calculating Loudness According to DIN45631 (ISO532B), Journal Acoustic Society Jpn (E), Vol. 12, Nr.1,
1991.
[17] S.J. Stevens, Procedure for Calculating Loudness : Mark VII, J. Acoust. Soc. Am., Vol. 33, Nr.11, pp.1577-1585, 1961.
[18] E. Zwicker, Procedure for Calculating Loudness of Temporally Variable Sounds, J. Acoust. Soc. Am., Vol. 62, Nr. 3, pp 675-681, 1977.
[19] German standard with amendment, Ref No. DIN 45631/A1.
[20] D. Siano, M.V. Prati, M. A. Costagliola, M.A. Panza, Evaluation of Noise Level Inside Cab of a Bi-Fuel Passenger Vehicle, WSEAS
Transactions on Applied and Theoretical Mechanics, ISSN / E-ISSN: 1991-8747 / 2224-3429, Volume 10, 2015, Art. #24, pp. 220-226.
Biography
M. A. Panza was born in Cosenza in 23/06/1986 Italy, and graduated in Mechanical Engineering at
the University of Calabria, Italy in 2012.She is currently a PhD Student in Industrial Engineering at
University of Naples Federico II, Italy, in collaboration with CNR - Istituto Motori, Italy. She is
dealing with acoustic and vibration in automotive field.
1023
ORIGINAL PAPER
Received: 26 October 2014 / Accepted: 18 August 2015 / Published online: 27 August 2015
# The Author(s) 2015. This article is published with open access at SpringerLink.com
Abstract
Purpose The paper addresses evolution of fuzzy systems for
core applications of automotive engineering.
Methods The presented study is based on the analysis of bibliography dedicated to fuzzy sets and fuzzy control for ground
vehicles. A special attention is given to fuzzy approaches used
in the following domains of automotive engineering: vehicle
dynamics control systems, driver and driving environment
identification, ride comfort control, and energy management
of electric vehicles.
Results The bibliographical analysis, supplemented with statistics of relevant research publications, has allowed to allocate the most important fuzzy application cases for each domain. In particular, it concerns driver identification, humanmachine interface, recognition of road conditions, and controllers of vehicle chassis and powertrain systems. It is found out
that fuzzy methods have the primary use most of all for tasks
requiring identification and forecasting procedures, especially
in conditions of limited informational space. Additional observation that can be also derived from the presented survey
points to reasonable integration of fuzzy technique with other
control engineering methods to improve the performance of
automotive control systems.
Conclusions In the aggregate the performed review indicates
that fuzzy computing can be considered as a versatile tool for
automotive engineering applications of different nature.
* Valentin Ivanov
valentin.ivanov@tu-ilmenau.de
1
1 Introduction
Fuzzy logic, among other computational intelligence
methods, attracts increased attention of engineers and researchers involved in the development of complex control
solutions for road vehicles and their subsystems. A possibility
to overcome various nonlinear models and to use intuitive
logical rules makes fuzzy control and fuzzy identification procedures useful tools by rapid system prototyping and on early
design stages. It is of particular relevance for intelligent vehicle functions and systems requiring the prediction of manoeuvres dynamics, identification of driving environment parameters and objects, and interaction with the driver. For the mentioned topics, the development of models of objects and relevant controllers of on-board systems can meet essential problems with feasibility, sufficient level of complexity and other
features demanded for real-time automotive applications.
These issues can be efficiently handled with the computational
intelligence methods, which are known as reasonable and
powerful tools in solving non-well-posed analytical problems.
Among different computational intelligence methods like neural networks, swarm intelligence computing et al. the fuzzy
logic can be of special relevance for intelligent automotive
systems. The reason is that the fuzzy logic has good applicability for identification tasks and pre-emptive control under
strong presence of data and model uncertainty.
The above-listed arguments are motivated the study
outlined in this paper. The main goal of the study is to give a
survey of existing fuzzy methods and systems that are finding
applications in the automotive area. Because the subject of the
study is characterized by a high degree of interdisciplinarity,
29 Page 2 of 10
the discussed review can be of particular interest for researchers working on topics of Ground Vehicle Engineering,
Control Engineering, and Computational Intelligence.
Specific objectives of the presented work can be formulated
as follows:
Define research topics that are most relevant to automotive fuzzy systems and tools;
Identify the main application cases for each defined topic;
Estimate development trends for each defined topic;
Analyse particular matters of automotive fuzzy systems
from viewpoint of (i) industrial applications and (ii) integration with other control techniques.
The results of the performed survey are given in next sections in accordance with the listed objectives. The article includes also extensive bibliography aimed at the representation
of various research schools around the world in context of the
discussed topics.
2 Review background
The history of fuzzy systems in vehicle-related areas begins
with several research works of founders of fuzzy logic dedicated to practical applications of fuzzy sets. An example is the
intersection controller that reduced junction-related delay of
vehicle driving proposed by Pappis and Mamdani in [1]. Later
Sugeno and Murakami in [2] and Sugeno and Nishida in [3]
have introduced the fuzzy realization of drivers logic for the
cases of an automated car parking system and handling the
desired vehicle trajectory. Then, from early 1990s, more and
more automotive systems with fuzzy logic are subjected both
to academic and industrial areas. This resulted in a considerable amount of publications and patents.
An analysis of the published studies and works allows to
allocate a number of domains, where fuzzy logic has found the
most acceptance. These domains can be conditionally called
as BDriver^, BVehicle Dynamics^, BRide comfort^, BElectric
vehicles^, and BDriving environment^. Results of analysis of
publications relevant to the fuzzy modeling and control of
automotive systems in the listed engineering areas are further
introduced in the paper.
The presented review uses the methodology proposed and
partially implemented in [4]. The analysis uses only Englishlanguage scientific publications found in Scopus abstract and
citation database of peer-reviewed literature (http://www.
elsevier.com/online-tools/scopus). The following limitations
of the analysis are accepted within the framework of this
paper: The works are analyzed in relation to the ground and
road vehicles only. Mobile robots and machines with nonholonomic constraints are excluded from the analysis due to
their less relevance to automotive applications.
The overview is accompanied by statistics of peerreviewed journal articles from two recent decades (dated
19942013) in each domain. (Conference papers are excluded
from statistical analysis to avoid the consideration of duplicated content. The elimination has been also done for those papers, which content was already published in another journal.
This is because it was observed that in particular cases the
same content is repeated with minor variations in several conference papers published by the same authors.)
Additional section of the paper will analyse the publications of SAE International (http://www.sae.org) as one of the
most recognized informational source worldwide for
automotive engineering and presenting mainly the studies,
which are providing experimental verification and close to
industrial application.
Page 3 of 10 29
29 Page 4 of 10
Methods of objective evaluation and control of the vehicle ride comfort [6668],
Control algorithms of suspension actuators and shock absorbers with adaptation to environmental and operational
conditions such as road roughness, vehicle dynamic variables and other [6973]
Special control strategies of electrorheological,
magnetorheological, and electrical dampers and specific
suspension elements [7477].
Page 5 of 10 29
29 Page 6 of 10
10 Conclusions
Analysis of research literature, especially results of published
experimental works, demonstrates that fuzzy methods have
solid background and good prospects for the implementation
Table 1
in automotive engineering applications. The following positions can be especially mentioned in this context:
Page 7 of 10 29
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
Open Access This article is distributed under the terms of the
Creative Commons Attribution 4.0 International License (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted
use, distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a link
to the Creative Commons license, and indicate if changes were made.
21.
22.
23.
References
Pappis CP, Mamdani EH (1977) A fuzzy logic controller for a
traffic junction. IEEE Trans Syst Man Cybernet SMC-7(10):
707717
2. Sugeno M, Murakami K (1984) Fuzzy parking control of model
car, Proc. IEEE Conf Decis Control, pp. 902903
3. Sugeno M, Nishida M (1985) Fuzzy control of model car. Fuzzy
Sets Syst 16:103113
4. Ivanov V (2001) Fuzzy methods in ground vehicle engineering :
state-of-the-art and advanced applications, in Proc. 8th
International Conference on Structural Dynamics, EURODYN
2011, Leuven, Belgium, paper MS19-270
5. Kramer U, Rohr G (1982) A model of driver behaviour.
Ergonomics 25(10):891907
24.
1.
25.
26.
27.
28.
29 Page 8 of 10
29.
Huh K, Kim J (2001) Active steering control based on the estimated tire forces. J Dyn Syst Meas Control Trans ASME 123(3):
505511
51. Jalali K, Uchida T, McPhee J, Lambert S (2013) Development of
an advanced fuzzy active steering controller and a novel method to
tune the fuzzy controller. SAE Int J Passenger Cars Electron Electr
Syst 6(1):241254
52. Basset M, Zimmer C, Gissinger GL (1997) Fuzzy approach to the
real time longitudinal velocity estimation of a FWD car in critical
situations. Veh Syst Dyn 27(56):477489
53. Hwang JK, Song CK (2005) Fuzzy estimation of vehicle speed
using an accelerometer and wheel sensors. Int J Automot Technol
6(4):359365
54. Cheli F, Sabbioni E, Pesce M, Melzi S (2007) A methodology for
vehicle sideslip angle identification: comparison with experimental data. Veh Syst Dyn 45(6):549563
55. Zeyada YF, Elbeheiry EM, Elarabi ME (2000) Fault diagnosis and
-estimation for integrated chassis control of vehicles in emergency maneuvers. J Eng Appl Sci 47(4):665681
56. Yoshimura T, Emoto Y (2003) Steering and suspension system of
a full car model using fuzzy reasoning and disturbance observers.
Int J Veh Auton Syst 1(34):363386
57. March C, Shim T (2007) Integrated control of suspension and
front steering to enhance vehicle handling. Proc IMechE Part D
J Automob Eng 221(4):377391
58. Song J (2012) Integrated control of brake pressure and rear-wheel
steering to improve lateral stability with fuzzy logic. Int J Automot
Technol 13(4):563570
59. Ergun M (2007) Estimation of friction coefficient of asphalt concrete road surfaces using the fuzzy logic approach. Can J Civ Eng
34(9):11101118
60. Shyrokau BN, Ivanov VG (2008) Alterable fuzzy sets in automotive control applications. Int J Model Identif Control 3(3):305317
61. Li L, Song J, Li H-Z, Shan D-S, Kong L, Yang CC (2009)
Comprehensive prediction method of road friction for vehicle dynamics control. Proc IMechE Part D J Automob Eng 223(8):9871002
62. Ivanov V, Shyrokau B, Augsburg K, Algin V (2010) Fuzzy evaluation of tyre-surface interaction parameters. J Terrramech 47(2):
113130
63. Ivanov V (2011) Fuzzy architecture of systems with alterable information: case study for tyreground friction estimators. Int J
Reliab Saf 5(3/4):398419
64. Yester JL, McFall RH (1992) Fuzzy logic controller for active
suspension. Proc Soc Automot Eng P-260:259270
65. Lin Y-J, Lu Y-Q (1993) Toward better ride performance of vehicle
suspension systems via intelligent control. Am Soc Mech Eng
Dyn Syste Control Div (Publ) DSC 48:105112
66. Jianghong Z, Long T (1994) An evaluation of comfort of a bus
seat. Appl Ergon 25(6):386392
67. Eslaminasab N, Biglarbegian M, Melek WW, Golnaraghi MF
(2007) A neural network based fuzzy control approach to improve
ride comfort and road handling of heavy vehicles using semiactive dampers. Int J Heavy Veh Syst 14(2):135157
68. Suzuki T, Takahashi M (2010) Robust active suspension control
for vibration reduction of passengers body. Trans JSME Part C
76(762):380387
69. Cherry AS, Jones RP (1995) Fuzzy logic control of an automotive
suspension system. IEE Proc Control Theory Appl 142(2):149160
70. Yoshimura T, Isari Y, Li Q, Hino J (1997) Active suspension of
motor coaches using skyhook damper and fuzzy logic control.
Control Eng Pract 5(2):175184
71. Caponetto R, Diamante O, Fargione G, Risitano A, Tringali D
(2003) A soft computing approach to fuzzy sky-hook control of
semiactive suspension. IEEE Trans Control Syst Technol 11(6):
786798
Sie W-T, Lian R-J, Lin B-F (2006) Enhancing grey prediction
fuzzy controller for active suspension systems. Veh Syst Dyn
44(5):407430
73. Kaldas MMS, alikan K, Henze R, Kkay F (2013) Rule
optimized fuzzy logic controller for full vehicle semi-active suspension. SAE Int J Passenger Cars Mech Syst 6(1):332344
74. Choi S-B, Park D-W, Suh M-S (2002) Fuzzy sky-ground hook
control of a tracked vehicle featuring semi-active
electrorheological suspension units. J Dyn Syst Meas Control
Trans ASME 124(1):150157
75. Giuclea M, Sireteanu T, Stancioiu D, Stammers CW (2004) Model
parameter identification for vehicle vibration control with
magnetorheological dampers using computational intelligence
methods. Proc IMechE Part I J Syst Control Eng 218(7):569581
76. Zheng L, Li YN, Baz A (2009) Fuzzy-sliding mode control of a
full car semi-active suspension systems with MR dampers. Smart
Struct Syst 5(3):261277
77. Lee S, Kim W-J (2010) Active suspension control with directdrive tubular linear brushless permanent-magnet motor. IEEE
Trans Control Syst Technol 18(4):859870, art. no. 5286231
78. Hogle RA, Bonissone PP (1994) Fuzzy algorithm for path selection in autonomous vehicle navigation. Proc IEEE Conf Decis
Control:898900
79. Freisleben B, Kunkelmann T (1993) Combining fuzzy logic and
neural networks to control an autonomous vehicle, 1993 I.E.
International Conference on Fuzzy Systems, pp. 321326
80. Blchl B, Tsinas L (1994) Automatic road following using fuzzy
control. Control Eng Pract 2(2):305311
81. Chakroborty P, Kikuchi S (1999) Evaluation of the general motors
based car-following models and a proposed fuzzy inference model. Transp Res Part C Emerg Technol 7(4):209235
82. Naranjo JE, Gonzlez C, Reviejo J, Garca R, De Pedro T (2003)
Adaptive fuzzy control for inter-vehicle gap keeping. IEEE Trans
Intell Transp Syst 4(3):132142
83. Qiao MS, Takeda H (1995) Learning algorithm of environmental
recognition in driving vehicle. IEEE Trans Syst Man Cybernet
25(6):917925
84. Stover JA, Hall DL, Gibson RE (1996) A fuzzy-logic architecture
for autonomous multisensor data fusion. IEEE Trans Ind Electron
43(3):403410
85. Goebel KF, Agogino AM (2001) Sensor validation and fusion for
automated vehicle control using fuzzy techniques. J Dyn Syst
Meas Control Trans ASME 123(1):145146
86. Rigatos G, Tzafestas S (2007) Extended Kalman filtering for
fuzzy modelling and multi-sensor fusion. Math Comput Model
Dyn Syst 13(3):251266
87. Kim SY, Choi HC, Won WJ, Oh SY (2009) Driving environment
assessment using fusion of in- and out-of-vehicle vision systems.
Int J Automot Technol 10(1):103113
88. Broggi A, Cerri P, Ghidoni S, Grisleri P, Jung HG (2009) A new
approach to urban pedestrian detection for automatic braking.
IEEE Trans Intell Transp Syst 10(4):594605, art. no. 5290131
89. Filev DP, Kolmanovsky I (2012) Markov chain modeling and onboard identification for automotive vehicles. Lect Notes Control
Inf Sci 418:111128
90. Bauer M, Tomizuka M (1996) Fuzzy logic traction controllers and
their effect on longitudinal vehicle platoon systems. Veh Syst Dyn
25(4):277303
91. Lee GD, Kim SW, Yim YU, Jung JH, Oh SY, Kim BS (1999)
Longitudinal and lateral control system development for a platoon
of vehicles, IEEE Conference on Intelligent Transportation
Systems, Proceedings, ITSC, pp. 605610
92. Kim K, Kim J, Huh KS, Yi K, Cho D (2006) A real-time multivehicle simulator for longitudinal controller design. Veh Syst Dyn
44(5):369386
Page 9 of 10 29
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
Pasquier M, Quek C, Toh M (2001) Fuzzylot: a novel selforganising fuzzy-neural rule-based pilot system for automated vehicles. Neural Netw 14(8):10991112
Dai X, Li CK, Rad AB (2005) An approach to tune fuzzy controllers based on reinforcement learning for autonomous vehicle control. IEEE Trans Intell Transp Syst 6(3):285293
Milans V, Villagr J, Prez J, Gonzlez C (2012) Low-speed
longitudinal controllers for mass-produced cars: a comparative
study. IEEE Trans Ind Electron 59(1):620628, art. no. 5759771
Prez J, Milans V, Godoy J, Villagr J, Onieva E (2013)
Cooperative controllers for highways based on human experience.
Expert Syst Appl 40(4):10241033
Kim HM, Dickerson J, Kosko B (1996) Fuzzy throttle and brake
control for platoons of smart cars. Fuzzy Sets Syst 84(3):209234
Naranjo JE, Reviejo J, Gonzlez C, Garca R, De Pedro T (2004)
A throttle and brake fuzzy controller: towards the automatic car.
Lect Notes Comput Sci (Incl Subseries Lec Notes Artif Intell Lect
Notes Bioinforma 2809:291301
Naranjo JE, Gonzlez C, Garca R, De Pedro T, Haber RE (2005)
Power-steering control architecture for automatic driving. IEEE
Trans Intell Transp Syst 6(4):406415
Ryu Y-W, Oh S-Y, Kim S-Y (2008) Robust automatic parking
without odometry using an evolutionary fuzzy logic controller.
Int J Control Autom Syst 6(3):434443
Chen C-Y, Feng H-M (2009) Hybrid intelligent vision-based carlike vehicle backing systems design. Expert Syst Appl 36(4):
75007509
Shiao Y, Yu C-J (2012) Automatic parking control for vehicles
with stereo vision. Adv Sci Lett 8:4954
Farrall SD, Jones RP (1993) Energy management in an automotive electric/heat engine hybrid powertrain using fuzzy decision
making, Proc. 1993 International Symposium on Intelligent
Control, pp. 463468
Cerruto E, Consoli A, Raciti A, Testa A (1994) Fuzzy logic based
efficiency improvement of an urban electric vehicle. IECON Proc
(Ind Electron Conf 2:13041309
Vermeiren L, Guerra TM, Paganelli G (1997) Application of fuzzy
set theory to electric car control. Cybern Syst 28(8):675693
Baumann BM, Washington G, Glenn BC, Rizzoni G (2000)
Mechatronic design and control of hybrid electric vehicles.
IEEE/ASME Trans Mechatron 5(1):5872
Schouten NJ, Salman MA, Kheir NA (2002) Fuzzy logic control
for parallel hybrid vehicles. IEEE Trans Control Syst Technol
10(3):460468
Ippolito L, Loia V, Siano P (2003) Extended fuzzy C-means and
genetic algorithms to optimize power flow management in hybrid
electric vehicles. Fuzzy Optim Decis Making 2(4):359374
Ahn H-S, Lee NS, Moon CW, Jeong G-M (2007) Fuel economy
improvement for fuel cell hybrid electric vehicles using fuzzy
logic-based power distribution control. Int J Automot Technol
8(5):651658
Solano Martnez J, Mulot J, Harel F, Hissel D, Pra M-C, John RI,
Amiet M (2013) Experimental validation of a type-2 fuzzy logic
controller for energy management in hybrid electrical vehicles.
Eng Appl Artif Intell 26(7):17721779
Yoshimura T, Uchida H, Nasu H, Hino J, Ueno R (1997) Traction
force control of an electric vehicle in 2WS-4WD mode using fuzzy
reasoning. Int J Veh Des 18(5):442454
Pusca R, Ait-Amirat Y, Berthon A, Kauffmann JM (2002) Fuzzy
logic based control for electric vehicle with four separate traction
drives. Proc IEEE Veh Technol Conf 4:20892097
Kim D-H, Kim J-M, Hwang S-H, Kim H-S (2007) Optimal brake
torque distribution for a four-wheel-drive hybrid electric vehicle
stability enhancement. Proc IMechE Part D J Automob Eng
221(11):13571366
29 Page 10 of 10
114.
115.
116.
117.
118.
119.
120.
121.
122.
123.
124.
125.
126.
127.
128.
129.
130.
Huang H (1999) Electrical two-speed propulsion by motor winding switching and its control strategies for electric vehicles. IEEE
Trans Veh Technol 48(2):607618
Sepe RB Jr, Miller JM (2002) Real-time collaborative experimentation via the internet: fuzzy efficiency optimization of a hybrid
electric vehicle starter/alternator. Int J Veh Auton Szstems 1(1):
6382
Ferreira AA, Pomilio JA, Spiazzi G, de Araujo Silva L (2008)
Energy management fuzzy logic supervisory for electric vehicle
power supplies system. IEEE Trans Power Electron 23(1):107
115
Quigley CP, Ball RJ, Jones RP (2000) Fuzzy modelling approach
to the prediction of journey parameters for hybrid electric vehicle
control. Proc IMechE Part D J Automob Eng 214(8):875885
Cunningham I, Burnham K (2013) Online use of the fuzzy transform in the estimation of electric vehicle range. Meas Control
(UK) 46(9):277282
Chau KT, Wu KC, Chan CC, Shen WX (2003) A new battery
capacity indicator for nickel-metal hydride battery powered electric vehicles using adaptive neuro-fuzzy inference system. Energy
Convers Manag 44(13):20592071
Albertus P, Couts J, Srinivasan V, Newman J (2008) A combined
model for determining capacity usage and battery size for hybrid
and plug-in hybrid electric vehicles. J Power Sources 183(2):771
782
Fleischer C, Waag W, Bai Z, Sauer DU (2013) On-line self-learning time forward voltage prognosis for lithium-ion batteries using
adaptive neuro-fuzzy inference system. J Power Sources 243:728
749
Hayashi K, Shimizu Y, Dote Y, Takayama A, Hirako A (1995)
Neuro fuzzy transmission control for automobile with variable
loads. IEEE Trans Control Syst Technol 3(1):4952
Halfmann C, Ayoubi M, Holzmann H (1997) Supervision of vehicles tyre pressures by measurement of body accelerations.
Control Eng Pract 5(8):11511159
Hissel D, Candusso D, Harel F (2007) Fuzzy-clustering durability
diagnosis of polymer electrolyte fuel cells dedicated to transportation applications. IEEE Trans Veh Technol 56(5 I):24142420,
art. no. 4305501
Kim MH, Lee S, Lee KC (2011) A fuzzy predictive redundancy
system for fault-tolerance of x-by-wire systems. Microprocess
Microsyst 35(5):453461
Mur S, Demichela M, Piccinini N (2006) Assessment of the risk
of occupational accidents using a Bfuzzy^ approach. Cognit
Technol Work 8(2):103112
Mon Y-J (2007) Airbag controller designed by adaptive-networkbased fuzzy inference system (ANFIS). Fuzzy Sets Syst 158(24):
27062714
Khayyam H, Nahavandi S, Hu E, Kouzani A, Chonka A, Abawajy
J, Marano V, Davis S (2011) Intelligent energy management control of vehicle air conditioning via look-ahead system. Appl
Therm Eng 31(16):31473160
Nopiah ZM, Junoh AK, Ariffin AK (2013) Noise annoyance
fuzzy index in passenger car cabin. Int J Veh Noise Vib 9(34):
216233
Farkas L, Moens D, Donders S, Vandepitte D (2012) Optimisation
study of a vehicle bumper subsystem with fuzzy parameters. Mech
Syst Signal Process 32:5960
132.
133.
134.
135.
136.
137.
138.
139.
140.
141.
142.
143.
144.
145.
146.
147.
REVIEW
Robert S. Bigler1
Received: 13 November 2015 / Accepted: 22 February 2016 / Published online: 16 April 2016
The Author(s) 2016. This article is published with open access at Springerlink.com
1 Introduction
The continued development of advanced driver assistance
systems (ADAS) in road vehicles is resulting in increasingly
complex interactions between driver and vehicle (Gordon
This work was supported by the UK Engineering and Physical
Sciences Research Council (EP/P505445/1) (studentship for Nash).
Christopher J. Nash
cn320@cam.ac.uk
David J. Cole
djc13@cam.ac.uk
123
92
Road
geometry
Driver
Road
preview
Preview
time step
Trajectory
planning
Steering
control
Motion
perception
Speed
control
Vehicle
motion
Steering
wheel angle
Pedal
forces
Preview horizon
Vehicle
Vehicle
disturbances
Fig. 1 Block diagram of tasks carried out while driving. The driver
must use their perceived information about the motion of the vehicle
and the upcoming road geometry to plan a desired trajectory and then
calculate the required steering wheel angle and pedal forces to achieve
this trajectory as closely as possible
vehicle motion is not used directly for generating the feedforward control action; however, the feedback loop is able to
correct for any discrepancies introduced by imperfections in
the drivers feedforward control. It has been found that without visual feedback during lane change or obstacle avoidance
manoeuvres drivers do not always initiate the return phase of
the manoeuvre, failing to steer back towards the target path
(Wallis et al. 2002; Cloete and Wallis 2009).
Modelling the driver mathematically has been the subject of research for many decades. Comprehensive reviews
are provided by Macadam (2003) and Plchl and Edelmann
(2007). Recent research has focussed on the application
of optimal control theory, using model predictive or linear
quadratic controllers that are able to preview the future road
path, as shown in Fig. 2, and calculate an optimal sequence
of control actions (Macadam 1981; Sharp and Valtetsiotis
2001; Peng 2002; Cole et al. 2006). This approach has been
extended to include neuromuscular dynamics (Pick and Cole
2007, 2008; Odhams and Cole 2009; Abbink et al. 2011;
Cole 2012) and to the control of nonlinear vehicle dynamics
(Ungoren and Peng 2005; Thommyppillai et al. 2009; Keen
and Cole 2011). Feedforward and feedback control are usually assumed to share a common objective function. Timings
and Cole (2014) synthesised independent feedforward and
feedback controllers to examine in more detail the robustness of the drivers control strategy to disturbances.
123
Previewed
displacements
93
Receptor
nonlinearities
Receptor
dynamics
Stimulus
= Delay sources
+
+
Sensors
Nerve
conduction
= Noise sources
Brain
Signal
processing
Spontaneous
neuron firing
Reflex
response
Sensory
integration
Muscles
+
Response
Neuromuscular
dynamics
Muscle
activation
Response
planning
Internal
model
+
Nerves
Neuromuscular
nonlinearities
Internal
model
errors
+
+
environment. The planned signals are sent to the muscles and shaped
by the neuromuscular dynamics to give a physical response. There are
various sources of time delays, shown by boxes, and noise, shown by
ovals
signals characterised by the dynamics of the sensory receptors (explored in Sect. 2). Sensory signals are then transmitted
through the nerves as electrical impulses caused by firing
neurons, with the firing rate encoding a frequency-modulated
signal (Carpenter 1984). Certain stimuli can elicit reflexive
responses which bypass the brain by activating motor neurons emerging from the spinal cord (Carpenter 1984).
There are physical and biochemical limitations to the
speed with which each of the processes shown in boxes in
Fig. 3 can be carried out; therefore, time delays are introduced
into the sensorimotor system. These delays are discussed
further in Sect. 3. In addition, noise is introduced due to
nonlinearities in the receptor and neuromuscular dynamics,
errors in the brains internal models and spontaneous firing
of neurons (Fernandez and Goldberg 1971). This means that
humans are unable to measure stimuli with perfect accuracy
or plan and execute an ideal response. It also results in thresholds below which stimuli cannot be perceived, as discussed
in Sect. 4.
Once the sensory signals are received in the brain, they
are processed in the sensory cortex in order to extract the
information from the encoded signals transmitted through
the nerves (Kandel et al. 2000). The information from the
different senses is then integrated to form a single representation of the surrounding environment, as explained further in
Sect. 5. Based on this, the physical response to the perceived
stimuli is planned using internal models of the human body
and the surrounding world (Wolpert and Ghahramani 2000).
123
94
The scope of this review is broad, and thus it is not possible to review every topic in great detail; each section could be
extended significantly. However, the aim of the review is to
give an overview of the key results from the literature, with
particular focus on motivating and informing further development of driver models incorporating human sensory system
dynamics. Both steering and speed control are considered
concurrently, since in many cases the sensory mechanisms
discussed are relevant for both control tasks. The main findings of the review are summarised and discussed in Sect. 6.
The review extends considerably an earlier review by Bigler
and Cole (2011).
2 Sensory dynamics
Various sensory systems are used by the driver to infer the
state of the vehicle and its surroundings. The main sensory
systems used in the control of vehicle speed and direction
are:
Visual: The visual system is the only means the driver
has of detecting the upcoming road geometry. The visual
system can also sense the motion of the vehicle relative
to the surrounding environment.
Vestibular: The vestibular organs are located within the
inner ear, and they sense rotations and translations of the
drivers head.
Somatosensory: Somatosensors include a wide range of
sensory organs which detect various states of the body,
such as contact pressure, temperature, limb position and
pain. They include proprioceptors which detect joint
angles, muscle lengths and tensions and their derivatives.
The following subsections give an overview of the published
literature on these three sensory systems. Other senses such
as hearing may also play a role but will not be discussed in
detail.
2.1 Visual system
Visual perception is the subject of significant research activity in psychology, neuroscience and biology. There is still
much to understand about how a human interprets the neural
signals received by the retina from a potentially complex
three-dimensional visual scene containing objects that might
be familiar or unfamiliar, and moving or stationary, with
a moving or stationary observer. The various processes
involved in visual perception are discussed in detail by Gibson (1950), Johansson (1975), Ullman (1979), Nakayama
(1985), Lappe et al. (1999) and Raudies and Neumann
(2012). Human visual perception is a complex, multi-layered
process, and for the purpose of driver modelling it is not nec-
123
Vanishing
point
95
FRO
Future
path point
Tangent
point
(a)
(b)
Fig. 4 Potential candidates for visual cues used while driving along a
a straight road and b a curved road. Optic flow patterns are shown by
dashed lines. Drivers have also been found to use other objects and the
road boundaries and centre line as visual references. Points such as the
tangent point and future path point have been identified as fixation
points for feedforward visual perception.
vi
s + vi
(1)
123
96
tangent point over the future path point or other nearby points
as a fixation point during driving (Mars 2008; Robertshaw
and Wilkie 2008; Lappi et al. 2013).
Eye tracking studies have found that drivers tend to focus
on a point around 12 s ahead of the vehicle on straight roads
(Land and Lee 1994; Donges 1978), and that their gaze tends
to move to an upcoming curve around 1 s before they steer in
that direction (Chattington et al. 2007; Land and Tatler 2001).
Drivers have also been found to make short look-ahead fixations, looking further along the road for short periods of time
(Lehtonen et al. 2013). While eye tracking instrumentation is
useful for determining the gaze direction of a driver, Land and
Lee (1994) noted that it does not necessarily indicate where
the driver is directing their attention, because the driver may
be using their peripheral vision to gather information about
road geometry away from the gaze point. Grunwald and Merhav (1976) and Land and Horwood (1995) both measured
driver performance with only certain parts of the road visible
and found that the full visual control task can be represented
by two viewing points, one near to the driver and one further
down the road. Land and Horwood (1995) found that performance was not degraded from the full visibility condition if
drivers could see a near point 0.53 s ahead and a distant point
0.93 s ahead.
Steen et al. (2011) reviewed many studies which proposed
one, two or multi-point preview models and concluded that
a two-point preview model was the most realistic, with one
point close to the driver and one more distant point. However, Sharp and Valtetsiotis (2001) used a shift register to
formulate a multi-point preview controller using visual information taken from a single preview point, suggesting that a
human driver in a moving vehicle could use memory to construct a multi-point image of the road geometry from data
sensed at just one or two discrete points. The use of linear
quadratic optimal control theory to calculate the gains on
multi-point road path geometry ahead of the vehicle shows
that the gains eventually tend to zero as the time ahead of
the vehicle increases. This indicates that looking beyond a
certain point might result in diminishing returns (Sharp and
Valtetsiotis 2001; Cole et al. 2006), with the time ahead of
the vehicle at which this occurs dependent on the dynamic
properties of the vehicle and the driver, and the amount of
control effort applied by the driver.
(1 + Toto1 s)
(1 + Toto2 s)(1 + Toto3 s)
(2)
123
97
K oto
Toto1
Toto2
Toto3
0.4
13.2
5.33
0.66
Hosman (1996)
0.4
0.5
0.016
0.4
10
0.016
0.0225
22.05
0.62
0.016
Magnitude (abs)
10 0
10 -1
10 -2
The semicircular canals consist of sets of three elliptical cavities which are each filled with fluid (Kandel et al. 2000).
Angular motion about any axis causes the fluid to move
within these cavities, causing deflections of small hair cells
which excite sensory cells. Early models of the SCCs were
based on considerations of the physical dynamics of the
organs. Steinhausen (1933) used observations of the motion
within the SCCs of fish to develop the torsion-pendulum
model. Young and Oman (1969) adapted this model to
include additional adaptation terms TSCCa to match trends
seen in experimental results. Fernandez and Goldberg (1971)
added a lead term TSCC1 , giving the transfer function:
TSCCa s
(1 + TSCCa s)
(1 + TSCC1 s)
(1 + TSCC2 s)(1 + TSCC3 s)
(3)
Phase (deg)
10 -3
90
45
0
-45
-90
10 -3
10 -2
10 -1
10 0
10 1
10 2
10 3
Frequency (rad/s)
Fig. 5 Bode plot for otolith transfer function with parameters from
different studies, given in Table 1. Input is acceleration
exponent, which is difficult to implement practically. Therefore, Hosman (1996) proposed a simplified version in the
same form as Eq. 2. Based on this and other research, Telban and Cardullo (2005) suggested parameters for a transfer
function in the form of Eq. 2, relating the specific force input
to a perceived specific force output. Soyka et al. (2011) used
a signal-in-noise model to find a transfer function for the
otoliths which optimised the fit to sensory threshold measurements (see Sect. 4 for more information). Suggested otolith
parameters from these studies are summarised in the remaining rows of Table 1. The gains K oto have been adjusted to
give comparable outputs, since the scaling of the output signal is arbitrary. Bode plots of the otolith transfer function
using the different parameters are compared in Fig. 5. For
a driving task, the mid-range frequencies (between around
101 and 101 rad/s) are the most important, and in this range
the otoliths exhibit a roughly proportional response to accelerations. There are differences in the details of the frequency
responses measured in different studies, which highlights the
difficulty in achieving repeatable results when using different
subjects, equipment and methodologies.
TSCCa s 2
(1 + TSCCa s)(1 + TSCC2 s)
(4)
123
98
Study
K SCC
TSCCa
TSCC1
TSCC2
TSCC3
5.73
80
0.049
5.70
0.005
Hosman (1996)
5.73
(80)
0.110
5.90
0.005
5.73
80
(0.060)
5.73
(0.005)
2.2
()
0.014
2.16
0.005
Parameters which the authors have suggested may be neglected are given in brackets
Magnitude (abs)
10 2
10 0
10 -2
Phase (deg)
10 -4
90
45
0
-45
10 -2
10 0
10 2
10 4
Frequency (rad/s)
Fig. 6 Bode plot for SCC transfer function with parameters from different studies, given in Table 2. Input is angular acceleration
why Telban and Cardullo (2005) suggested the SCCs measure angular velocity rather than acceleration. In contrast to
the otolith dynamics, the agreement between the different
studies is much higher. This may be because these studies
based their work on similar models of the physical dynamics
of the SCCs, although the transfer function found from sensory thresholds (Soyka et al. 2012) also agrees well with the
others at mid-range frequencies.
2.3 Somatosensors
During driving, the information provided by the visual and
vestibular systems is complemented by the response of various receptors of the somatosensory system (Kandel et al.
2000). A particular group of receptors provide proprioception, which is the sensing of joint angles and movements
and muscle displacements and forces. These receptors are
particularly important in allowing the driver to sense the
angle and torque of the steering wheel, which can be used
by experienced drivers to sense the characteristics of the
contact between the tyre and the road. Proprioceptors are
also used to sense the displacements and forces of the foot
123
Muscle spindles are sensors which detect the length and rate
of change of length of the muscles. They produce two separate signals, one dependent on muscle velocity and length
(type Ia afferent) and one dependent on muscle length only
(type II afferent) (Kandel et al. 2000). An empirical linear
model of the muscle spindle response, based on measurements taken in cats, was formulated by Poppele and Bowman
(1970), with the Ia and II afferent responses to muscle displacements given by:
s(s + 0.44)(s + 11.3)(s + 44)
(s + 0.04)(s + 0.816)
(s + 0.44)(s + 11.3)
HII (s) =
(s + 0.816)
HIa (s) =
(5)
(6)
More complicated nonlinear models have also been developed which can predict the afferent responses accurately
under a wide variety of conditions (Maltenfort and Burke
2003; Mileusnic et al. 2006).
2.3.2 Golgi tendon organs
Golgi tendon organs (GTOs) respond to the forces in the
muscles. They share a nerve with the Ia afferent response of
the muscle spindles, giving a response known as a type Ib
afferent (Kandel et al. 2000). A linear model of the GTOs
was first proposed by Houck and Simon (1967), again based
on measurements in cats. Their model was stated as a transfer
99
(7)
3 Time delays
As shown in Fig. 3, there are various ways in which delays are
introduced between sensory stimuli being applied to a driver
and the drivers control response being measured. Delay
sources include receptor dynamics, nerve conduction, neural
processing and neuromuscular dynamics. Various techniques
have been used in the literature to measure delays in human
response to sensory stimulation. The simplest of these is to
apply a stimulus and measure the time taken for a physical
response (such as pressing a button) to be recorded. Some
studies have used more sophisticated methods of applying stimuli, such as galvanic vestibular stimulation (GVS)
which bypasses the vestibular organs by applying an electrical stimulus directly to the nerves (Fitzpatrick and Day
2004). Other methods have been used to detect responses at
other points in the process, such as measuring the VOR to
identify the reflexive delay, using magnetoencephalography
(MEG, Hmlinen et al. 1993) or electroencephalography
(EEG) to measure electrical impulses within the brain or
using electromyography (EMG) to record electrical activity
in the muscles.
When interpreting sensory time delays measured in different studies using different techniques, it is important to
consider which of the delay components shown in Fig. 3 are
included in the measurement in each case. The aim of this
section is to use results from the literature to estimate the
total delay between stimulus and response for each sensory
system. However, it can be difficult to separate the effects of
pure time delays from lags due to the dynamics of the sensors and muscles and the time taken for signals to rise above
noise levels (Soyka et al. 2013). Nevertheless, results from
the literature can be used to find an approximate estimate
of the order of magnitude of time delays in human sensory
systems.
EMG has been used to measure the response of the muscle spindles to applied muscle stretches, finding delays of
2530 ms for the Ia afferent and 40 ms for the II afferent
(Matthews 1984). Bigler (2013) combined these with measured nerve conduction delays (Trojaborg and Sindrup 1969;
Kandel et al. 2000) to give delays of 34 ms and 48 ms for the
Ia and II afferents. As the Ib afferent response of the GTOs
123
4 Perception thresholds
Due to limits of human sensory organs and noise caused
by spontaneous neuron firing, sensory systems have thresholds below which stimuli cannot be perceived. Perception
thresholds are defined as the smallest stimulus which can
be detected, and these are commonly measured by asking
subjects to distinguish something about the stimulus, such
as its direction. In reality, these thresholds are generally not
precise, but a smooth transition from 0 to 100 % probability
of detection over a range of values. This cumulative probability distribution is known as a psychometric function
(Boring 1917) and is often modelled as a cumulative normal distribution. Variations on the updown method (Levitt
1970) are commonly used to measure perception thresholds,
and depending on the method used the thresholds measured
correspond to different probabilities of detection, generally
between 65 and 80 %.
The just noticeable difference (JND) is defined as the
smallest change in amplitude from a reference stimulus
which is required before the difference between the two stimuli is noticed. From experiments on the perception of lifted
masses, Weber (1834) found that the JND in mass was proportional to the reference mass. This result has been found to
be applicable for many perceptual systems and has become
known as Webers law with the constant of proportionality known as the Weber fraction. Figure 7 shows how the
123
JND
100
Weber
Fraction
1
Perception
Threshold
Stimulus Intensity
Fig. 7 JNDs for a stimulus following Webers law. Webers law states
that JNDs increase linearly with stimulus intensity. The constant of
proportionality is known as the Weber fraction, and the y-intercept is
the perception threshold
101
Method 1
Method 2
Perceived Amplitude
Perceived Amplitude
Stimulus
Amplitude
Stimulus
Amplitude
SDN
Stimulus
Sensor
Transfer
Function
SDN
Gain
+
+
AN
Gain
+
+
are found in this way for the different senses using sensory
threshold measurements from the literature. These measurements have all been taken under passive unimodal conditions;
therefore, since thresholds have been found to increase under
active or multimodal conditions (see Sect. 4.3) the noise
amplitudes found in this section can be considered to be
lower bounds. For each sensory system, the signal-in-noise
model of Soyka et al. (2011, 2012) has been used to identify the additive noise amplitudes using two different transfer
functions: (i) a published sensor transfer function from considerations of the sensory dynamics and (ii) a sensor transfer
function optimised to fit threshold data. It is unclear which
of the two transfer functions is more appropriate for driver
modelling. The parameters derived from sensory threshold
measurements may describe the behaviour at low amplitudes
better; however, they may not completely match the dynamic
behaviour of the sensory system. Noise amplitudes are given
in units with a * symbol at the end, to indicate that the noise is
added to the stimuli filtered by the sensory transfer functions.
4.2.1 Visual thresholds
Perceived
Stimulus
123
0.01
Measurements
Model (dynamics)
Model (thresholds)
0.008
0.006
0.004
0.002
0
10
10
100
101
Frequency (Hz)
(a)
Yaw
102
0.25
Measurements
Model (dynamics)
Model (thresholds)
0.2
0.15
0.1
0.05
0
10
10
100
101
Frequency (Hz)
(b)
Sway
so the absolute values of the thresholds may not be at the correct scale.
Model fits are shown using two different transfer functions, one from
a simple model of the sensory dynamics of the visual system and one
which was optimised to fit the measured thresholds
123
103
0.1
Benson et al. (1986) X
Benson et al. (1986) Y
Soyka et al. (2011) X/Y
Soyka et al. (2009) X/Y
0.08
0.06
0.04
Kingma (2005) X
Kingma (2005) Y
Model (dynamics)
Model (thresholds)
0.02
0
10
10
100
101
Frequency (Hz)
0.12
Benson et al. (1989) exp. 1
Benson et al. (1989) exp. 2
Grabherr et al. (2008)
Soyka et al. (2012)
Model (dynamics)
Model (thresholds)
0.1
0.08
0.06
0.04
0.02
0
10
10
100
101
Frequency (Hz)
Fig. 11 Lateral (Y) and longitudinal (X) acceleration thresholds measured in several different studies, compared with models found from
the dynamics of the otoliths and from threshold measurements
of brain responses, and the solid line shows the results using
parameters optimised by Soyka et al. (2011) to fit the measured thresholds. The threshold model was found using the
results of Soyka et al. (2011) only, whereas the noise level
for the dynamics transfer function was optimised to fit the
whole data set. The noise levels found were 0.038 m/s2 *
for the dynamics transfer function and 0.015 m/s2 * for the
thresholds transfer function. The transfer function found
from the threshold measurements fits the results much better
than the transfer function found from the sensory dynamics,
as the corner frequency at which the thresholds plateau is too
low for the dynamics transfer function.
Naseri and Grant (2012) measured JND values for sinusoidal accelerations at 0.4 and 0.6 Hz with varying amplitudes. The results were found to fit Webers law well,
although a dependence on frequency was also seen. A Weber
fraction of 5 % was found for the measurements taken at
0.4 Hz, whereas a value of 2 % was found for the measurements taken 0.6 Hz.
In interpreting the results of experiments which measure
thresholds of whole body motion, the possibility of multimodal stimuli should be considered. For example, in the case
of sinusoidal angular velocity imposed on the test subject, the
semicircular canals and various somatosensors may be stimulated simultaneously. Multimodal thresholds and sensory
integration are discussed in Sects. 4.3 and 5.
studies using the updown method are shown in Fig. 12. The
data all follow a similar trend, with a fairly low amount of
scatter compared to the otolith results. Predicted thresholds
are also shown using the signal-in-noise model of Soyka et al.
(2012), based on the transfer function given in Eq. 3. The
solid line was found using parameters optimised by Soyka
et al. (2012) to fit the threshold measurements, and the dotted
line was found using the parameters suggested by Telban and
Cardullo (2005) for the SCCs, choosing the noise level to fit
the measured threshold parameters as well as possible. Both
sets of SCC parameters are given in Table 2. The noise levels
found were 0.025 rad/s* for the thresholds transfer function
and 0.023 rad/s* for the dynamics transfer function. Both
models fit the results well, although the model which was
optimised to fit the threshold results matches more closely as
expected.
JNDs for angular velocity perception have been measured
by Mallery et al. (2010) and dos Santos Buinhas et al. (2013),
finding Weber fractions of 3 and 13 %, respectively. The
difference between these values may be a result of the fact
that Mallery et al. (2010) measured JNDs at larger amplitudes
than dos Santos Buinhas et al. (2013). Mallery et al. (2010)
also found that the gradient (JND/amplitude) was higher at
low amplitudes, and suggested a power law should be used
rather than Webers law. However, it is debatable whether
JNDs for the SCCs should follow a power law, when most
other sensory systems have been found to follow Webers
law.
123
104
0.03
0.025
0.02
0.015
0.01
0.005
0
10
10
100
101
10 2
Fig. 13 Thresholds for the perception of steering wheel angular displacement, measured by Bigler (2013). These results could be used to
infer somatosensor noise levels; however, assumptions would have to
be made about the integration of signals from the two muscle spindle
afferents and the GTOs
123
The studies summarised in Sect. 4.2 were all targeted at measuring thresholds of a single stimulus in isolation, during
passive conditions where the subject was concentrating on
the stimulus. However, sensory stimuli which occur during
driving are very different to the stimuli applied in these controlled studies, so these results may not be directly applicable
to driving tasks. Stimuli in driving tasks are perceived under
active rather than passive conditions, and there are several
stimuli being perceived at once. Therefore, the indifference
threshold [threshold in the presence of other stimuli (Groen
et al. 2006)] should determine the limits of perception during
driving.
By asking subjects to perform a secondary control task
in a separate motion axis, it has been found that increasing
the mental load on subjects causes an increase in perception thresholds (Hosman and Van Der Vaart 1978; Samji and
Reid 1992). It should be noted that in both of these studies the
subjects were still actively concentrating on perceiving the
motion cues as well as completing the secondary task. Due
to the equivalence of translational accelerations and shifts in
the gravity vector, the brain can easily be fooled into misinterpreting the two types of motion. Groen and Bles (2004)
and Pretto et al. (2014) found that presenting subjects with
visual cues simulating a translational acceleration while they
were undergoing rotational motion caused threshold of perception of the rotation to increase by factors of 56. Pretto
et al. (2014) also measured thresholds during an active control task and found that they increased by factors up to 4 for
some subjects, but didnt change at all for others. The participants whose thresholds did not increase during the active
driving task reported higher levels of immersion in the simulation, indicating that the sense of realism of the simulation
was linked to participants ability to perceive the motion cues
accurately.
Pitch and roll thresholds have been measured with masking vertical motion cues, finding a significant linear increase
in pitch and roll thresholds with vertical amplitude (Zaichik
et al. 1999; Rodchenko et al. 2000). In contrast to these studies, Valente et al. (2006) found no significant effect of vertical
motion amplitude on pitch rate thresholds. In this study, the
pitch and vertical motion were applied at the same frequency,
which may have caused the motion cues to be perceived as
coherent, making it easier to detect the pitch cues.
Groen et al. (2006) analysed the data of Groen and Bles
(2004) and showed that indifference thresholds for pitch rotation in the presence of visual longitudinal cues follow the
same frequency response as the perception thresholds measured in passive conditions, but are increased by a constant
gain. They used this result to hypothesise that the presence of
additional sensory stimuli scales perception thresholds by a
constant gain, without affecting the frequency response. This
105
one-to-one
line
PMC
CZW
Coherence
Zone
123
106
5 Sensory integration
The sensory systems described in Sect. 2 provide the central nervous system (CNS) with measurements (or sensory
cues) which can be used to estimate vehicle states while
driving. However, these measurements are shaped by the sensor dynamics and also contain additive and signal-dependent
noise (as described in Sect. 4). The CNS must therefore carry
out sensory integration to give a single estimate of the vehicle states from the noisy, filtered information received from
each of the sensors.
In a real-world driving scenario, the driver will be presented with coherent sensory information. Any discrepancies
between information from the different sensors are due
to sensory noise, or incomplete information available to a
particular sensor. However, in some situations the information presented to the different senses may be incoherent or
biased, in which case the driver may use a different integration strategy. This is particularly relevant for motion
in virtual environments, where the visual, vestibular and
somatosensory information presented to the driver may not
all accurately reflect the real-world stimuli. An overview
of methods and results from investigations of sensory integration in a variety of virtual environments (not specific to
driving) is given by Campos and Blthoff (2012). The following subsections build on this, focusing in more depth on
results which suggest how information from the sensory systems summarised in Sect. 2 may be integrated during driving.
5.1 Integration of coherent sensory measurements
The simplest model of sensory integration is a linear weighting of the estimates from different sensory systems (Hosman
and Stassen 1999). Appropriate weightings can be found
using sensory experiments; however, the scope of models
with fixed weightings is likely to be limited. For many
sensory systems, the CNS has been found to integrate measurements using statistically optimal methods (Ernst and
Banks 2002; Oru et al. 2003; Butler et al. 2010; Seilheimer
et al. 2014). These methods are based on Bayes theorem
(Bayes 1763), which relates the a posteriori probability
P( I| S) of condition I given observation S to the probability
P( I| S) of observation S given condition I, the a priori probability P( I), and the observation probability P( S) (which is
usually assumed uniform):
P( I| S) =
123
P( S| I)P( I)
P( S)
(8)
i
1/ 2
wi Si with wi = i 2
j 1/ j
(9)
(10)
107
(11)
(12)
Fig. 15 Maximum likelihood integration of two biased sensory channels. The probability distributions of the sensors S1 and S2 (shown by
the solid lines) are both biased, i.e. their means do not correspond to
the true value of I. Using MLE causes this bias to carry through to the
combined estimate given by the dashed line. Also note that the sensor
with the largest variance does not necessary have the largest bias
123
108
2011; Butler et al. 2010, 2014) or if the conflicting information is presented in different motion axes (Kaliuzhna et al.
2015). de Winkel et al. (2015) found that over half of subjects
integrated visual and inertial heading information regardless
of the size of the bias. However, other studies have found
evidence of various strategies for reducing bias in perceived
signals (Krding et al. 2007; Landy et al. 1995; Burge et al.
2010; Zaidel et al. 2011).
When presented with two different sensory cues, the CNS
must decide whether or not they are coherent (originating
from the same source). If they are coherent, the difference
between them can be assumed to be a result of stochastic
variations and the cues can be combined using MLE. If not,
the cues should be treated separately, treating the situation as
a cue conflict. Krding et al. (2007) proposed a model using
Bayes rule to decide whether or not two cues are coherent
based on a prior describing the likelihood of the cues coming from the same source. They validated the model using
experimental results; however, Seilheimer et al. (2014) noted
that Krding et al. (2007) did not vary the reliability of the
cues, so it is still uncertain whether their model is valid in
all cases. A similar Bayesian model incorporating priors was
proposed by Knill (2007). They found that the weight applied
to a cue shrunk as the size of the conflict increased, but it
did not decrease to zero. However, other studies have found
that under some circumstances humans will veto a cue that
does not fit with the other sensory measurements (Girshick
and Banks 2009; Landy et al. 1995).
Ghahramani et al. (1997) proposed an additional stage of
cue calibration before cues are fully integrated, where the
difference between the estimates is reduced. The values of
estimates S1 and S2 are calibrated by adding S1 and S2 ,
given by:
S1 = C1 ( S2 S1 )
S2 = C2 ( S1 S2 )
(13)
123
Ci = 0 for the visual channel and Ci = 1 for the other channel. Alternatively, Ghahramani et al. (1997) proposed that
the calibration stage follows a similar weighting structure
to the integration process, based on reliability. He hypothesised that each calibration constant C1 is proportional to the
variance i2 of the sensory estimate Si . Burge et al. (2010)
tested this model in an experiment on visual-haptic estimation of slant and reported strong evidence in favour of this
reliability-based calibration.
Reliability-based calibration does not make physical sense
as a method for reducing sensory bias, however, as the reliability of a cue is independent of its bias (Ernst and Luca 2011).
A sensory estimate could have low variance (high reliability)
and high bias, or a high variance (low reliability) but low bias.
For example, in Fig. 15, the cue with the higher variance has
the lower bias. Testing a cue veto model of integration of
biased sensory estimates, Girshick and Banks (2009) found
that the vetoed cue was not necessarily the cue with the highest variance. Zaidel et al. (2011) compared reliability-based
calibration with fixed-ratio calibration, where the calibration
constants Ci were assumed to be learned from past experience. Fixed-ratio calibration fitted their results better than
reliability-based calibration, with higher weighting placed on
the visual estimate. The sum of the calibration constants C1
and C2 was found to be less than 1, so full internal consistency was not achieved. Zaidel et al. (2011) also explained
how the presence of fixed-ratio calibration could cause erroneous indications of reliability-based calibration to appear
using the methods of Burge et al. (2010). It therefore seems
that with biased sensory information humans may use fixed
calibration constants based on past experience, rather than
change the weightings based on cue reliability.
Linear cue calibration for visualvestibular integration has
been observed in several studies although, as with coherent measurements, there is disagreement about which sense
is more highly weighted. Visual dominance was found by
Rader et al. (2011), whereas the vestibular system was found
to dominate by Harris et al. (2000). Ohmi (1996) found that
visual cues dominated when conflicts were small, but vestibular cues dominated when conflicts were large. Zacharias and
Young (1981) found that vestibular cues dominated visual
cues at higher frequencies, so the dominant sensory system
may depend on the frequency content of the tasks carried
out.
Experimental studies have shown that when a consistent
conflict is observed between the visual and vestibular systems, the perceived motion will eventually drift towards the
visual estimate (Ishida et al. 2008). van der Steen (1998) proposed a model of the optokinetic influence, where the visual
estimate attracts the vestibular estimate over a transient
period, modelling the onset of visual self-motion (vection).
This is modelled by passing the difference between the visual
and vestibular estimates through a low-pass filter given by:
HOK (s) =
109
1
1 + OK s
(14)
Vestibular
cue
Vestibular
Dynamics
0.5
+
HP Filter
Vestibular
Dynamics
(Internal Model)
Visual
cue
0.5+
+
+
Perceived
motion
Visual
Dynamics
123
110
Input
Weber
fraction (%)
Sensor
delay (ms)
0.0011 (rad/s*)
711
100560
0.032 (m/s*)
711
100560
1050
100560
11
100560
0.015 (m/s2 *)
25
5440
0.025 (rad/s*)
313
5440
Transfer function
Noise
Transfer function
Noise
0.810
s + 0.810
Visual feedback
Yaw angular
velocity
0.0013 (rad/s*)
Visual feedback
Lateral velocity
0.035 (m/s*)
Visual feedback
Longitudinal
velocity
0.810
s + 0.810
Visual
feedforward
Target path
Preview model
Otoliths
Acceleration
SCCs
Angular velocity
Muscle spindles
(Type Ia)
Arm muscle
displacement
0.4(1 + 10s)
(1 + 5s)(1 + 0.016s)
5.73(80s 2 )
(1 + 80s)(1 + 5.73s)
s(s + 0.44)(s + 11.3)(s + 44)
(s + 0.04)(s + 0.816)
Muscle spindles
(Type II)
Arm muscle
displacement
GTOs
0.038 (m/s2 *)
0.023 (rad/s*)
0.0225(1 + 22.05s)
(1 + 0.62s)(1 + 0.016s)
2.2s(1 + 0.014s)
(1 + 2.16s)(1 + 0.005s)
10
>34
(s + 0.44)(s + 11.3)
(s + 0.816)
10
>48
10
>34
For the key sensory systems involved in driving, transfer functions between the input stimulus and the sensory response are given, either from
considerations of the sensory dynamics or from perception threshold measurements. Noise levels have been calculated from sensory threshold
measurements, as well as Weber fractions showing how thresholds increase with stimulus amplitude. Estimates of sensory delays are also included
6 Discussion
Key results from the literature on human sensory dynamics
have been presented in Sects. 2 to 5. In this section, these
results are summarised and discussed with a view to understanding and modelling driver steering and speed control.
Results for the human sensory systems which are most relevant to driver modelling are summarised in Table 3. Transfer
functions are presented which have either been found from
models of the sensory dynamics and measurements of brain
activity or inferred from sensory threshold measurements.
Using the transfer functions found from sensory threshold
data may give more accurate results near the limits of perception; however, they may not capture all of the dynamic
behaviour of the sensory system.
Noise magnitudes have been inferred from sensory threshold measurements using the signal-in-noise model of Soyka
et al. (2011, 2012). These were found from passive threshold measurements taken for one sensory stimulus at a time;
however, thresholds have been found to increase in active
conditions and in the presence of other sensory stimuli by
factors between 1.5 and 6 (Hosman and Van Der Vaart 1978;
Samji and Reid 1992; Zaichik et al. 1999; Rodchenko et al.
2000; Groen and Bles 2004; Valente Pais et al. 2012). This
means that the noise values shown in Table 3 should be considered as lower bounds. Most sensory systems have been
found to approximate Webers law, with JNDs increasing
with stimulus amplitude; therefore, Weber fractions have
123
7 Conclusion
The results summarised in this literature review give an
insight into various different sensory systems, and how they
can be used to model driver control behaviour. Sensory transfer functions have been studied extensively, and there is little
disagreement between different studies. Sensory integration
is reasonably well understood under normal conditions; however, there is little agreement on how humans cope with
conflicting sensory information. Studies have shown that
111
sensory thresholds increase under active and multimodal conditions, but further research is necessary to determine how
and why this happens. Time delays also increase during multimodal conditions; however, it is not clear whether they vary
during active control tasks. There is a great deal of scope
for improvement in the available knowledge on human sensory perception during active control tasks, so future research
should focus in this area. It is hoped that the information in
this review will prove useful in developing more sophisticated driver steering and speed control models which take
account of the drivers sensory dynamics.
Acknowledgments The authors wish to thank the anonymous reviewer
who provided many useful suggestions for improving the paper, particularly regarding visual perception.
Open Access This article is distributed under the terms of the Creative
Commons Attribution 4.0 International License (http://creativecomm
ons.org/licenses/by/4.0/), which permits unrestricted use, distribution,
and reproduction in any medium, provided you give appropriate credit
to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made.
References
Abbink Da, Mulder M, Van Der Helm FCT, Mulder M, Boer ER
(2011) Measuring neuromuscular control dynamics during car following with continuous haptic feedback. IEEE Trans Syst Man
Cybern Part B Cybern 41(5):12391249. doi:10.1109/TSMCB.
2011.2120606
Authi CN, Mestre DR (2012) Path curvature discrimination: dependence on gaze direction and optical flow speed. PloS One
7(2):e31,479. doi:10.1371/journal.pone.0031479
Aw ST, Todd MJ, Halmagyi GM (2006) Latency and initiation of the
human vestibuloocular reflex to pulsed galvanic stimulation. J
Neurophysiol 96(2):925930. doi:10.1152/jn.01250.2005
Barber MJ, Clark JW, Anderson CH (2003) Neural representation
of probabilistic information. Neural comput 15(8):18431864.
doi:10.1162/08997660360675062
Barnett-Cowan M, Nolan H, Butler JS, Foxe JJ, Reilly RB, Blthoff
HH (2010) Reaction time and event-related potentials to visual,
auditory and vestibular stimuli. J Vis 10(7):e1400. doi:10.1167/
10.7.1400
Barnett-Cowan M (2013) Vestibular perception is slow: a review. Multisens Res 26(4):387403. doi:10.1163/22134808-00002421
Barnett-Cowan M, Harris LR (2009) Perceived timing of vestibular
stimulation relative to touch, light and sound. Exp Brain Res
198(23):221231. doi:10.1007/s00221-009-1779-4
Bayes T (1763) An essay towards solving a problem in the doctrine of
chances. Philos Trans R Soc Lond 53:370418. doi:10.1098/rstl.
1763.0053
Benson AJ, Spencer MB, Stott JR (1986) Thresholds for the detection
of the direction of whole-body, linear movement in the horizontal
plane. Aviat Space Environ Med 57:10881096
Benson AJ, Hutt EC, Brown SF (1989) Thresholds for the perception of
whole body angular movement about a vertical axis. Aviat Space
Environ Med 60:205213
Bigler RS, Cole DJ (2011) A review of mathematical models of
human sensory dynamics relevant to the steering task. In: Iwnicki
S, Goodall R, Mei TX (eds) The International Association for
123
112
Vehicle System Dynamics, Manchester Metropolitan University,
Manchester, UK
Bigler RS (2013) Automobile driver sensory system modeling. Ph.d.
thesis, Cambridge University
Boer ER (1996) Tangent point oriented curve negotiation. In: Proceedings of conference on intelligent vehicles, IEEE 617:712. doi:10.
1109/IVS.1996.566341
Boer ER (1999) Car following from the drivers perspective. Transp
Res Part F Traffic Psychol Behav 2(4):201206. doi:10.1016/
S1369-8478(00)00007-3
Borah J, Young LR, Curry RE (1988) Optimal estimator model for
human spatial orientation. Ann N Y Acad Sci 545(1):5173.
doi:10.1111/j.1749-6632.1988.tb19555.x
Boring E (1917) A chart of the psychometric function. Am J Psychol
28(4):465470
Bottoms DJ (1983) The interaction of driving speed, steering difficulty and lateral tolerance with particular reference to agriculture.
Ergonomics 26(2):123139. doi:10.1080/00140138308963324
Bremmer F, Lappe M (1999) The use of optical velocities for
distance discrimination and reproduction during visually simulated self motion. Exp Brain Res 127(1):3342. doi:10.1007/
s002210050771
Bronstein AM, Hood JD (1986) The cervico-ocular reflex in normal
subjects and patients with absent vestibular function. Brain Res
373(12):399408. doi:10.1016/0006-8993(86)90355-0
Browning NA, Grossberg S, Mingolla E (2009) A neural model of how
the brain computes heading from optic flow in realistic scenes.
Cogn Psychol 59(4):320356. doi:10.1016/j.cogpsych.2009.07.
002
Burge J, Girshick AR, Banks MS (2010) Visual-haptic adaptation is
determined by relative reliability. J Neurosci Off J Soc Neurosci
30(22):77147721. doi:10.1523/JNEUROSCI.6427-09.2010
Butler JS, Smith ST, Campos JL, Blthoff HH (2010) Bayesian integration of visual and vestibular signals for heading. J Vis 10(11):23.
doi:10.1167/10.11.23
Butler JS, Campos JL, Blthoff HH, Smith ST (2011) The role of
stereo vision in visual-vestibular integration. Seeing Perceiving
24(5):453470. doi:10.1163/187847511X588070
Butler JS, Campos JL, Blthoff HH (2014) Optimal visualvestibular integration under conditions of conflicting intersensory
motion profiles. Exp Brain Res 233(2):587597. doi:10.1007/
s00221-014-4136-1
Campos JL, Blthoff HH (2012) Multimodal integration during selfmotion in virtual reality. In: Murray MM, Wallace MT (eds) The
neural bases of multisensory processes, chap 30. CRC Press, Boca
Raton
Carpenter RHS (1984) Neurophysiology, 4th edn. Arnold, London
Carriot J, Jamali M, Cullen KE (2015) Rapid adaptation of multisensory
integration in vestibular pathways. Front Syst Neurosci 9(59):15.
doi:10.3389/fnsys.2015.00059
Chattington M, Wilson M, Ashford D, Marple-Horvat DE (2007) Eyesteering coordination in natural driving. Exp Brain Res 180(1):1
14. doi:10.1007/s00221-006-0839-2
Clark JJ, Yuille AL (1990) Data fusion for sensory information processing systems. Kluwer Academic Publishers, Norwell
Cloete SR, Wallis G (2009) Limitations of feedforward control in
multiple-phase steering movements. Exp Brain Res 195(3):481
487. doi:10.1007/s00221-009-1813-6
Cole DJ, Pick AJ, Odhams A (2006) Predictive and linear
quadratic methods for potential application to modelling driver
steering control. Veh Syst Dyn 44(3):259284. doi:10.1080/
00423110500260159
Cole DJ (2012) A path-following driver-vehicle model with neuromuscular dynamics, including measured and simulated responses to
a step in steering angle overlay. Veh Syst Dyn 50(4):573596.
doi:10.1080/00423114.2011.606370
123
113
Gu Y, Angelaki DE, Deangelis GC (2008) Neural correlates of
multisensory cue integration in macaque MSTd. Nat Neurosci
11(10):12011210. doi:10.1038/nn.2191
Hmlinen M, Hari R, Ilmoniemi RJ, Knuutila J, Lounasmaa
OV (1993) Magnetoencephalography-theory, instrumentation, and
applications to noninvasive studies of the working human brain.
Rev Mod Phys 65(2):413497. doi:10.1103/RevModPhys.65.413
Harris LR, Jenkin M, Zikovitz DC (2000) Visual and non-visual cues in
the perception of linear self-motion. Exp Brain Res 135(1):1221.
doi:10.1007/s002210000504
Heerspink H, Berkouwer W, Stroosma O, van Paassen MM, Mulder M,
Mulder B (2005) Evaluation of vestibular thresholds for motion
detection in the SIMONA research simulator. In: AIAA modeling and simulation technologies conference and exhibit, American
Institute of Aeronautics and Astronautics, Reston, Virigina, p
e6502. doi:10.2514/6.2005-6502
Herrin GD, Neuhardt JB (1974) An empirical model for automobile
driver horizontal curve negotiation. Hum Factors 16(2):129133.
doi:10.1177/001872087401600203
Hosman RJ (1996) Pilots perception and control of aircraft motions.
Ph.d. Thesis, Delft University of Technology
Hosman RJ, Stassen H (1999) Pilots perception in the control of aircraft motions. Control Eng Pract 7(11):14211428. doi:10.1016/
S0967-0661(99)00111-2
Hosman RJ, Van Der Vaart JC (1978) Vestibular models and thresholds
of motion perception: results of tests in a flight simulator. Tech.
rep. Delft University of Technology, Delft
Houck JA, Simon W (1967) Responses of Golgi tendon organs to forces
applied to muscle tendon. J Neurophysiol 30(6):14661481
Ishida M, Fushiki H, Nishida H, Watanabe Y (2008) Self-motion perception during conflicting visualvestibular acceleration. J Vestib
Res Equilib orientat 18:267272
Johansson G (1975) Visual motion perception. Sci Am 232(6):7688.
doi:10.1038/scientificamerican0675-76
Johns TA, Cole DJ (2015) Measurement and mathematical model of a
drivers intermittent compensatory steering control. Vehicle system dynamics (In press) NVSD110,074. doi:10.1080/02699931.
2011.628301
Jonik P, Valente Pais AR, van Paassen MM, Mulder M (2011) Phase
coherence zones in flight simulation. In: AIAA modeling and simulation technologies conference, American Institute of Aeronautics
and Astronautics, Reston, Virigina, August, p e6555. doi:10.2514/
6.2011-6555
Kaliuzhna M, Prsa M, Gale S, Lee SJ, Blanke O (2015) Learning to
integrate contradictory multisensory self-motion cue pairings. J
Vis 15(1):e10. doi:10.1167/15.1.10
Kalman RE (1960) A new approach to linear filtering and prediction
problems. J Basic Eng 82(1):3545. doi:10.1115/1.3662552
Kandel E, Schwartz J, Jessell T (2000) Principles of neural science, 4th
edn. McGraw-Hill, New York
Kandil FI, Rotter A, Lappe M (2009) Driving is smoother and more
stable when using the tangent point. J Vis 9(2009):11.111.11.
doi:10.1167/9.1.11
Kandil FI, Rotter A, Lappe M (2010) Car drivers attend to different gaze targets when negotiating closed vs. open bends. J Vis
10(2010):24.124.11. doi:10.1167/10.4.24
Kawakami O, Kaneoke Y, Maruyama K, Kakigi R, Okada T, Sadato N,
Yonekura Y (2002) Visual detection of motion speed in humans:
spatiotemporal analysis by fMRI and MEG. Hum Brain Mapp
118:104118. doi:10.1002/hbm.10033
Keen SD, Cole DJ (2011) Application of time-variant predictive control
to modelling driver steering skill. Veh Syst Dyn 49(4):527559.
doi:10.1080/00423110903551626
Keen SD, Cole DJ (2012) Bias-free identification of a linear modelpredictive steering controller from measured driver steering behav-
123
114
ior. IEEE Trans Syst Man Cybern Part B Cybern 42(2):434443.
doi:10.1109/TSMCB.2011.2167509
Kingma H (2005) Thresholds for perception of direction of linear acceleration as a possible evaluation of the otolith function. BMC Ear
Nose Throat Disord. doi:10.1186/1472-6815-5-5
Knill DC (2007) Robust cue integration: a Bayesian model and evidence
from cue-conflict studies with stereoscopic and figure cues to slant.
J Vis 7(7):e5. doi:10.1167/7.7.5
Koenderink JJ (1986) Optic flow. Vis Res 26(1):161179. doi:10.1016/
0042-6989(86)90078-7
Kondoh T, Yamamura T, Kitazaki S, Kuge N, Boer ER (2008) Identification of visual cues and quantification of drivers perception
of proximity risk to the lead vehicle in car-following situations. J
Mech Syst Transp Logist 1(2):170180. doi:10.1299/jmtl.1.170
Krding KP, Beierholm U, Ma WJ, Quartz S, Tenenbaum JB, Shams
L (2007) Causal inference in multisensory perception. PloS One
2(9):e943. doi:10.1371/journal.pone.0000943
Lam K, Kaneoke Y, Gunji A, Yamasaki H, Matsumoto E, Naito T,
Kakigi R (2000) Magnetic response of human extrastriate cortex
in the detection of coherent and incoherent motion. Neuroscience
97(1):110. doi:10.1016/S0306-4522(00)00037-3
Land MF (1998) The visual control of steering. In: Harris LR, Jenkin M
(eds) Vision and action, Cambridge University Press, Cambridge,
UK, pp 163180
Land MF, Horwood J (1995) Which parts of the road guide steering?
Nature 377(6547):339430. doi:10.1038/377339a0
Land MF, Lee DN (1994) Where we look when we steer. Nature
369(6483):742744. doi:10.1038/369742a0
Land MF, Tatler BW (2001) Steering with the head. The visual strategy
of a racing driver. Current Biol CB 11(15):12151220
Landy MS, Maloney LT, Johnston EB, Young M (1995) Measurement
and modeling of depth cue combination: in defense of weak fusion.
Vis Res 35(3):389412
Lappe M, Bremmer F, Van Den Berg aV (1999) Perception of selfmotion from visual flow. Trends Cogn Sci 3(9):329336. doi:10.
1016/S1364-6613(99)01364-9
Lappi O, Lehtonen E, Pekkanen J, Itkonen T (2013) Beyond the tangent
point: gaze targets in naturalistic driving. J Vis 13(13):11. doi:10.
1167/13.13.11
Legge GE, Campbell F (1981) Displacement detection in human vision.
Vis Res 21(2):205213. doi:10.1016/0042-6989(81)90114-0
Lehtonen E, Lappi O, Kotkanen H, Summala H (2013) Look-ahead
fixations in curve driving. Ergonomics 56(1):3444. doi:10.1080/
00140139.2012.739205
Levitt H (1970) Transformed updown method in psychoacoustics. J
Acoust Soc Am 49(2b):467477. doi:10.1121/1.1912375
Ljung L (1999) System identification: theory for the user, 2nd edn.
Prentice Hall, Upper Saddle River
Lot R, Dal Bianco N (2015) Lap time optimisation of a racing go-kart.
Veh Syst Dyn 3114(January):121. doi:10.1080/00423114.2015.
1125514
Macadam CC (1981) Application of an optimal preview control for
simulation of closed-loop automobile driving. IEEE Trans Syst
Man CybernSMC 11(6):393399
Macadam CC (2003) Understanding and modeling the human
driver. Veh Syst Dyn 40(13):101134. doi:10.1076/vesd.40.1.
101.15875
MacNeilage PR, Banks MS, Berger DR, Blthoff HH (2007) A
Bayesian model of the disambiguation of gravitoinertial force
by visual cues. Exp Brain Res 179(2):263290. doi:10.1007/
s00221-006-0792-0
Mallery RM, Olomu OU, Uchanski RM, Militchin VA, Hullar TE
(2010) Human discrimination of rotational velocities. Exp Brain
Res 204(1):1120. doi:10.1007/s00221-010-2288-1
123
115
Scarfe P, Hibbard PB (2011) Statistically optimal integration of biased
sensory estimates. J Vis 11(7):e12. doi:10.1167/11.7.12
Seilheimer RL, Rosenberg A, Angelaki DE (2014) Models and
processes of multisensory cue combination. Curr Opin Neurobiol
25:3846. doi:10.1016/j.conb.2013.11.008
Sharp RS, Valtetsiotis V (2001) Optimal preview car steering control.
Veh Syst Dyn Suppl 35:101117
Shinar D, Mcdowell ED, Rockwell TH (1977) Eye movements in curve
negotiation. Hum Factors J Hum Factors Ergon Soc 19(1):6371.
doi:10.1177/001872087701900107
Soyka F, Robuffo Giordano P, Beykirch KA, Blthoff HH (2011)
Predicting direction detection thresholds for arbitrary translational acceleration profiles in the horizontal plane. Exp Brain Res
209(1):95107. doi:10.1007/s00221-010-2523-9
Soyka F, Giordano PR, Barnett-Cowan M, Blthoff HH (2012) Modeling direction discrimination thresholds for yaw rotations around
an earth-vertical axis for arbitrary motion profiles. Exp Brain Res
220(1):8999. doi:10.1007/s00221-012-3120-x
Soyka F, Blthoff HH, Barnett-Cowan M (2013) Temporal processing of
self-motion: modeling reaction times for rotations and translations.
Exp Brain Res 228(1):5162. doi:10.1007/s00221-013-3536-y
Soyka F, Blthoff HH, Barnett-Cowan M (2015) Integration of semicircular canal and otolith cues for direction discrimination during eccentric rotations. Plos One 10(8):e0136,925. doi:10.1371/
journal.pone.0136925
Soyka F, Teufel H, Beykirch KA, Robuffo Giordano P, Butler JS,
Nieuwenhuizen FM, Blthoff HH (2009) Does jerk have to be considered in linear motion simulation? In: Proceedings of the AIAA
modeling and simulation technologies conference, pp 13811388.
doi:10.2514/6.2009-6245
Steen J, Damveld HJ, Happee R, van Paassen MM, Mulder M (2011) A
review of visual driver models for system identification purposes.
In: 2011 IEEE international conference on systems, man, and
cybernetics, pp 20932100. doi:10.1109/ICSMC.2011.6083981
Steinhausen W (1933) ber die Beobachtung der Cupula in den Bogengangsampullen des Labyrinths des lebenden Hechts. Pflgers
Archiv fr die Gesamte Physiologie des Menschen und der Tiere
232(1):500512. doi:10.1007/BF01754806
Tabak S, Collewijn H, Boumans LM, Van Der Steen J (1997) Gain
and delay of human vestibulo-ocular reflexes to oscillation and
steps of the head by a reactive torque helmet. Acta Oto-laryngol
117(6):785795. doi:10.3109/00016489709114203
Telban RJ, Cardullo F (2005) Motion cueing algorithm development:
human-centered linear and nonlinear approaches. NASA technical
report. Nasa Langley Research Center, Hampton, VA, USA
Thommyppillai M, Evangelou S, Sharp RS (2009) Car driving at the
limit by adaptive linear optimal preview control. Veh Syst Dyn
47(12):15351550. doi:10.1080/00423110802673109
Timings JP, Cole DJ (2013) Minimum maneuver time calculation using
convex optimization. J Dyn Syst Meas Control 135(3):e031015.
doi:10.1115/1.4023400
Timings JP, Cole DJ (2014) Robust lap-time simulation. Proc Inst Mech
Eng Part D J Automob Eng 228(10):12001216. doi:10.1177/
0954407013516102
Todorov E (2005) Stochastic optimal control and estimation methods adapted to the noise characteristics of the sensorimotor system. Neural Comput 17(5):10841108. doi:10.1162/
0899766053491887
Trojaborg W, Sindrup EH (1969) Motor and sensory conduction in
different segments of the radial nerve in normal subjects. J Neurol
Neurosurg Psychiatry 32(4):354359
Ullman S (1979) Interpretation visual motion. Massachusetts Inst of
Technology Pr, Oxford
Ungoren A, Peng H (2005) An adaptive lateral preview
driver model. Veh Syst Dyn 43(4):245259. doi:10.1080/
00423110412331290419
123
116
Vaitl D, Mittelstaedt H, Saborowski R, Stark R, Baisch F (2002) Shifts
in blood volume alter the perception of posture: Further evidence
for somatic graviception. Int J Psychophysiol 44(1):111. doi:10.
1016/S0167-8760(01)00184-2
Valente Pais AR, Mulder M, van Paassen MM, Wentink M, Groen
EL (2006) Modeling human perceptual thresholds in self-motion
perception. In: AIAA modeling and simulation technologies
conference and exhibit, American Institute of Aeronautics and
Astronautics, Reston, Virigina, p e6626. doi:10.2514/6.2006-6626
Valente Pais AR, van Paassen MM, Mulder M, Wentink M (2010b)
Perception of combined visual and inertial low-frequency yaw
motion. In: AIAA modeling and simulation technologies conference, American Institute of Aeronautics and Astronautics, Reston,
Virigina, August, pp 110. doi:10.2514/6.2010-8093
Valente Pais AR, van Paassen MM, Mulder M, Wentink M (2011)
Effect of performing a boundary-avoidance tracking task on the
perception of coherence between visual and inertial cues. In: AIAA
modeling and simulation technologies conference, American Institute of Aeronautics and Astronautics, Reston, Virigina, August, p
e6324. doi:10.2514/6.2011-6324
Valente Pais AR, van Paassen MM, Mulder M, Wentink M (2010a) Perception coherence zones in flight simulation. J Aircr 47(6):2039
2048. doi:10.2514/1.C000281
Valente Pais AR, Pool DM, de Vroome AM, van Paassen MM, Mulder
M (2012) Pitch motion perception thresholds during passive and
active tasks. J Guid Control Dyn 35(3):904918. doi:10.2514/1.
54987
Valko Y, Lewis RF, Priesol aJ, Merfeld DM (2012) Vestibular labyrinth
contributions to human whole-body motion discrimination. J Neurosci 32(39):13,53713,542. doi:10.1523/JNEUROSCI.2157-12.
2012
van der Steen HFAM (1998) Self-motion perception. Ph.d. thesis, Delft
University of Technology, The Netherlands
Vaseghi S (2005) Advanced digital signal processing and noise reduction, vol 9, 3rd edn. Wiley, Chichester. doi:10.1002/0470094966
Wallis G, Chatziastros A, Blthoff HH (2002) An unexpected role for
visual feedback in vehicle steering control. Curr Biol 12(4):295
299. doi:10.1016/S0960-9822(02)00685-1
Weber E (1834) Annotationes anatomicae et physiologicae. CF Koehler,
Leipzig
Wentink M, Correia Grcio BJ, Bles W (2009) Frequency dependence
of allowable differences in visual and vestibular motion cues in a
simulator. In: AIAA modeling and simulation technologies conference, American Institute of Aeronautics and Astronautics, Reston,
Virigina, August, p e6248. doi:10.2514/6.2009-6248
Wierwille WW, Casali JG, Repa BS (1983) Driver steering reaction
time to abrupt-onset crosswinds, as measured in a moving-base
driving simulator. Hum Factors 25(1):103116
Wolpert DM, Diedrichsen J, Flanagan JR (2011) Principles of sensorimotor learning. Nat Rev Neurosci 12:739751. doi:10.1038/
nrn3112
Wolpert DM, Ghahramani Z (2000) Computational principles of movement neuroscience. Nat Neurosci Suppl 3:12121217. doi:10.
1038/81497
Wright WG, DiZio P, Lackner JR (2005) Vertical linear self-motion
perception during visual and inertial motion: more than weighted
summation of sensory inputs. J Vestib Res Equilib Orientat
15(4):185195
123
1. Introduction
Abbas and Bell (1994) outlined the modelling approach used in system dynamics (SD) and listed
12 advantages of the approach compared to traditional transport modelling. In particular, they
suggested the approach would be well suited to strategic policy analysis and as a support tool for
decision-making. In essence, transportation systems are complex, they often involve a number
of different stakeholders or agents which results in feedbacks with different time lags between
the responses of each type of user. System dynamics models offer a whole system approach to
transport planning and with this different perspective the importance of these feedbacks and lagged
responses can be demonstrated to policy makers. The SD platforms also offer specialised tools,1
which aid in the calibration of models, optimisation of policies and improve ease of use through
flight simulators which all contribute to the understanding of the whole underlying system.
System dynamics was initially developed by Forrester from MIT in the 19501960s (Forrester
1958) and is a powerful methodology developed from system theory, information science, organisational theory, control theory, tactical decision-making, cybernetics and military games. It uses
a standard causal loop approach to develop qualitative models of a system which may be used to
develop dynamic hypotheses before a quantitative stock-flow model is developed. Early applications were in business management but over the past few decades it has been applied to other areas,
including government policy, healthcare, the automobile industry and urban studies (Sterman
2000). The application of causal loop diagrams (which set out the causal links between concepts)
may be used to bring out the mental models (how people think a system works) of different
stakeholders and therefore help remove any barriers to implementation of a given policy. System
Email:
s.p.shepherd@its.leeds.ac.uk
84
S.P. Shepherd
dynamics approaches are becoming increasingly used in a hierarchical manner which allows systems and policies to interact across space and time. The holistic approach is well suited to the
transport problems we now face. Using different modelling approaches can also produce significantly different outcomes on the design of policies. System dynamics can bring in other modelling
structures and help explore the real drivers of future demand as well as explain how to change
user perception and behaviour. Models can be built with stakeholders input and then used in the
form of games or flight simulators for policy learning.
The aim of this review is to set out which areas of transportation research have applied SD
since the paper by Abbas and Bell, and in doing so highlight whether the studies have made
use of the qualitative causal loop approach, the quantitative stock-flow modelling approach and
which insights were possible (if any) over and above a more traditional modelling approach. The
approach taken was to review only peer-reviewed journal articles between 1995 and 2013. Whilst
there are many conference papers applying SD approaches to transport problems, the element of
peer-review was used to check for quality. A limited number of key word search terms including,
transport, transportation, SD, modelling were used and after filtering out via reading the abstracts
over 50 papers are included in the review. The papers are categorised by area of application in
what follows while the first section sets out some basics of the SD approach for those who are
unfamiliar with it. The paper then concludes with a discussion about future application areas.
Figure 1.
85
Figure 2. The bath tub metaphor for stock and flow with integral equation.
While qualitative models are useful in describing the structure of a system and a dynamic
hypothesis, most decision makers then wish to see some quantitative results. Here the approach is
based on linking differential equations but is presented to the user in terms of stocks and flows
via a stock-flow diagram which keeps the model transparent and easy to understand.
Stocks are accumulations and are represented by rectangles suggesting a box to hold the content.
Flows may be inflow to a stock or an outflow from a stock and are represented by pipes with
valves controlling the rate of flow into or out of a stock. Again taking the example from Sterman
(2000), the hydraulic or bathtub metaphor is used to explain the concept and mathematics behind
the simplest building block of one stock. Figure 2 shows the bathtub where the stock or level of
water accumulates over time with inflow controlled by the tap and outflow by the plug. The system
dynamic representation below shows the symbols used in VENSIM, one of the software platforms
used for developing SD models (others use very similar symbols). Underlying the symbols is the
mathematical notation which shows how the stock is the integral of inflowoutflow starting with
an initial level of stock. While the bath tub metaphor is easy to follow, it is also easy to see how
the stock could represent other systems such as population with births and deaths being the inflow
and outflow from the stock of people.
86
S.P. Shepherd
Table 1.
Application area
Number of papers
12
6
5
13
10
8
54
As stocks and flows are linked to other stocks and flows, a system structure develops and the
system is described by a series of linked non-linear differential equations. Like all modelling,
developing a good model is more like an art and part of the learning process is to develop
the model along with stakeholders. Where this is not possible, then simple concept models are
often useful to explain endogenous system behaviour and some system dynamicists prefer the
use of small models over the more complex and detailed ones see for example the discussion
in Ghaffarzadegan, Lyneis, and Richardson (2011) who discuss the benefits of using small SD
models when addressing public policy issues.
The following sections review over 50 journal papers which have applied SD to a transportation
problem. The papers have been categorised as in Table 1 which shows the category and number
of papers included in the review. Each section does not review all papers in detail but does draw
out examples of good practice which demonstrate the application of CLD or stock flow models
which provide something different to the more traditional transport modelling approach either in
terms of insight or coverage of the problem. As this is a review it is not feasible to go into the
detailed mathematics behind the stock-flow models, but readers may find this material within the
papers quoted.
87
This increase in vehicle miles driven may then impact on the scrappage loop R1, increasing
the demand for new vehicles earlier which increases production emissions. Finally, the authors
argue that the increased demand for high efficiency vehicles could via loop B1 (the demandprice effect), increase the price of such vehicles at least in the short term which goes against
the intended impact of the subsidy. Whilst the actual outcome depends on the relative strengths
of these loops or interactions, using such a diagram is useful in communicating the potential
unintended consequences or policy resistance which may occur. The authors also discuss how
policies which increase the cost per mile such as carbon taxes, may achieve synergy with such
subsidies via the CLDs.
88
S.P. Shepherd
Figure 4. Principal feedbacks conditioning willingness to consider new platforms and expected modes of
behaviour.
Source: Struben and Sterman (2008).
Moving on to the stock-flow models, Struben and Sterman (2008) develop a framework for modelling the uptake of alternate fuel vehicles (AFVs) which consists of three main elements: a fleet
turnover or stock model, a discrete choice model of the purchase decision and a social/technology
diffusion process (Figure 4). This process extends the Bass diffusion concept to include the impacts
of word of mouth, marketing and social exposure to the new vehicles. It allows for a development
of the choice set or willingness to consider the option over time including the effect of forgetting
about the new vehicle types which then permits a failing market solution. This is an example of
where SD models can bring something different to the process of policy assessment the possibility of an initial uptake of new technology which then fails and which has been observed in
practice when subsidies for compressed natural gas (CNG) vehicles were removed in Canada and
New Zealand (Flynn 2002). Whilst using an example of adoption of waste recycling, Ulli-Beer
et al. (2010) describe the general structures required to model acceptance dynamics involving
paradigm shift involving endogenous preference and value change, these structures are useful in
modelling acceptance and/or rejection of new technology in many areas.
Other studies incorporate similar diffusion structures but add other policy dimensions such as
regulation, including manufacturer responses, Walther et al. (2010), add other indicators such
as impact on fuel duty revenues (Shepherd, Bonsall, and Harrison 2012) and the co-evolution
of infrastructure for re-fuelling and extending to the cases of natural gas, Jannsen et al. (2006)
and hydrogen fuel cell vehicles (Meyer and Winebrake 2009; Kohler et al. 2010; Park, Kim, and
Lee 2011). Meanwhile others have investigated the role of strategic niche management, Kwon
89
(2012), the implication of policies on distributional issues, Harrison and Shepherd (2014) and
applied similar diffusion models to the case of electric motorcycles in Taiwan (Trappey et al.
2012). In terms of impact of policies, most studies find that the uptake is not affected greatly
by subsidies but more by regulation and infrastructure. However, as both Struben and Sterman
(2008) and Shepherd, Bonsall, and Harrison (2012) point out, the results may be more sensitive
to the assumed strength of the word of mouth or marketing effects rather than to changes in the
technical attributes of the alternate vehicles. This ability to bring in soft issues such as impacts
of social exposure and to quickly demonstrate the sensitivity of results to assume parameters are
another strength of the SD approach.
4. Supply chain management with transportation
Tako and Robinson (2012) provide a full review of discrete event and SD applications in the
logistics and supply chain context between 1996 and 2006. A total of 38 papers from 127 used
the SD approach which has a strong presence in the analysis of supply chains due to the natural
ability to represent the well-known bullwhip effect.2 Of these 38 papers only a few investigate
the impacts of supply chain management on the related transport system as shown in Table 1.
Sterman (2000) developed a generic SD model of the basic stock management structure which
explains the sources of oscillation, amplification and phase lag which are observed in real supply chains. Georgiadis, Vlachos, and Iakovou (2005) use these structures to construct a holistic
model by linking single echelon models with similar structures to represent different elements
of the supply chain. They apply the model to a case study of a fast food chain in Greece to
investigate capacity planning under dynamic growth assumptions and the question of when to
increase fleet size rather than lease extra capacity to meet delivery needs. The model is used to
optimise the number of company owned trucks while minimising total transportation costs. Potter
and Lalwani (2008) investigate a similar problem. Variation in daily demands is made worse by
demand amplification or the bullwhip effect and this can make it difficult for hauliers to judge
investments in their fleet compared with how much to sub-contract. The paper confirms previous
qualitative assessments in that demand amplification impacts adversely on transport costs and
performance/efficiency. However, Potter et al. are also able to identify exceptions when vehicle
capacity is just less than average demand, where an increase in demand amplification can improve
transport performance. This is due to spare capacity being available within the transport system;
however, this strategy comes with penalties incurred as extra vehicles are needed to deal with
average demand.
Disney, Potter, and Gardner (2003), Otto and Kotzab (2003) and Wilson (2007) use the SD
approach to investigate the potential for dealing with the bullwhip effect by using the vendor
managed inventory approach. Figure 5 shows a typical supply chain where times to respond and
estimates of demand for orders vary along the chain. Under the VMI approach the warehouse now
acts as a distribution centre and the tier 1 supplier controls the order process from direct estimates of
demand from the customer base. Information flows in the traditional approach between retailer and
warehouse and warehouse and tier 1 supplier are effectively replaced by this direct connection to
the customer which both improves accuracy of demand forecasts and reduces delays in the system.
The rest of the supply chain is as in the traditional model. Wilson (2007) investigated transportation
disruptions at each point along the supply chain. The most critical point was found to be between
the tier 1 supplier and the warehouse, and this was therefore seen as the most important point in
the supply chain and as the place to start with risk management strategies to mitigate the effect
of transportation disruptions. The simple approach to management of inventory by changing the
information flows was shown to reduce demand amplification and as Wilson (2007) shows VMI
can be used to protect against severe transportation disruptions. VMI protects the retailer against
90
Figure 5.
S.P. Shepherd
Supply chain management structure with VMI connection (adapted from Wilson 2007).
the disruption and they do not suffer a peak in inventory, sharing information essentially smooths
out the disruptive effects. The application of SD to this type of problem including transportation
was a natural extension of the more traditional SD literature on supply chain or stock management.
5. Highway maintenance/construction
There are five papers which deal with the highway maintenance and construction problem. Chasey,
de La Garza, and Drew (2002) and Fallah-Fini et al. (2010) model the impact of deferred maintenance on a highway system. Figure 6 adapted from Fallah-Fini et al. shows the aggregate feedback
loops between deterioration and maintenance. The pavement condition deteriorates due to load
91
factors and climate, etc. The balancing loop B1 (maintenance fix) shows that budget allocated
to maintenance operations increases the highway improvement rate which then decreases the
area of highway under distress which then balances or reduces the desired maintenance budget.
The reinforcing loop R1 (accelerated deterioration), includes the effect of maintenance budget
shortfall which causes a delay in maintenance which increases the highway deterioration rate
counter-acting the balancing loop. Whilst the physics of road deterioration and maintenance are
complex in nature, the corresponding feedback structure is relatively simple. Fallah-Fini et al.
developed a dynamic micro-level simulation model of highway deterioration and renewal processes. The model is calibrated with data from eight road sections in Virginia. The model is then
linked with an optimisation module to improve maintenance operations. The analysis results in
different priority setting schemes that improve on current maintenance practices at both the project
and network level. The approach suggests moving towards less costly preventative maintenance
rather than more expensive (deferred) corrective maintenance should bring in benefits to the system as a whole. Again a move away from the traditional approach to the problem provides insight
and promotes alternative strategies.
Friedman (2006) however, questions the preventative maintenance policy. Friedman models the
effect of road conditions on accident development and uses data to demonstrate that the mental
model behind the policy of maintaining better roads to reduce accidents could in most cases
be increasing accidents due to increased speeds and polishing effects. A more complex mental
and stock-flow model is developed which better fits the evidence of increased accidents with
increased pavement service index after maintenance. This demonstrates the importance of setting
boundaries when viewing problems and the importance of evidence to support the mental models
of basic decisions. Friedman questions the current mental model held within the road authority
and demonstrates the importance of accounting not only for empirical evidence but also for the
reasoning by decision-makers who develop their own mental models of short term consequences
where data are missing.
The other two papers in this area relate to policy in China. Hang and Li (2011) develop a model
to investigate the potential role of truck weight regulation policies for a region in China. While
it includes a fleet turn-over model, it does not make much use of other SD features. However,
the paper claims to bring insight which other models do not e.g. that adjustment to a rigid policy
(which requires a change in trucks) needs time and the disruption causes more problems to society
than a more moderate approach.
Xu, Mashayekhib, and Saeed (1998) develop a SD model to investigate the cost overrun problem
in Chinese highway construction projects. The existing policy was described as maximising
quantity over quality in terms of building many roads at low cost with an average life of only
15 years compared with 40 years for Western Europe. The model shows that as demand for
roads grows then the pressure to complete on time and allocation of budget between capital and
maintenance costs meant that quality was reduced as a result of the system itself. This was due
to the funding mechanism for new roads as revenues come from completed projects via the road
users. This meant that there was a pressure to build faster with lower quality so that more revenue
could be generated to fund more roads. A new equilibrium of project starts evolves but with a lower
quality. The study showed that this problem can be overcome with loan financing and increases in
user charges which releases budget constraints and hence pressures on the construction process
to deliver on time (but with reduced quality).
6. Strategic policy at urban, regional and national levels
This section includes studies of high level or strategic policy at the urban, regional, and national
level. Firstly, there are a number of models which look at implications for the structure of cities
92
S.P. Shepherd
and regions as the economy, population, migration, infrastructure and land use interact with
transportation. A sub-set of these build on the tradition of land use interaction models (LUTI). A
feature of land use transport interaction is that these two systems operate on different time scales.
Transport users may respond relatively quickly to changes in transport policy or costs, typically
within months while the land-use system includes a significant degree of inertia. This is due to the
fact that land use in urban areas depends on physical structures such as buildings and transport
infrastructure. This time dependence between the systems makes it ideal for a SD approach.
Despite this most LUTI models are based on the traditional approach in transport modelling i.e.
the notion of equilibrium and are static in nature. These equilibrium based models are used to
forecast changes in transport demand and land use 20 or 30 years hence with no real connection
from the current conditions to those future projections. The pathway to the future state is therefore
unknown.
Pfaffenbichler, Emberger, and Shepherd (2010) and Pfaffenbichler (2011) introduce the concepts underlying the MARS LUTI model. They cover validation and transferability between cities
and provide some examples of the model applied in practice. The model has now been applied in
more than 20 cities world-wide and has been used as a training tool for planners and practitioners
in Asia and in optimising strategic transport policies for a range of cities/regions. The model is
seen as a departure from the norm in transport planning yet has proved to be a useful tool at the
design stage. In particular, the model was first of all designed using a causal loop analysis. One
example concerns the link between planning decisions about infrastructure (road capacity) and the
symptom of congestion. A simplified view of the CLD reported in Pfaffenbichler (2011) is shown
in Figure 7 along-side one of the common two loop Archetypal Structures from Wolstenholme
(2003) which represents short-term fixes which fail. The left figure shows the common archetype
where a short-term fix to a problem then fails due to some unintended consequence which causes
the problem symptom to grow after some delay. Usually the positive reinforcing effects are outside
of the system boundary considered. The right-hand side shows the case for congestion which in
the past has been solved by planners with increased capacity which then increases the attractiveness of car use which in turn increases congestion again. Pfaffenbichler (2011) provides a more
detailed version of this concept and demonstrates that real-world behaviour follows this pattern
with evidence of traffic flows as infrastructure is developed for cases in Austria over a period
Figure 7.
Fixes that fail archetype and congestion capacity car use problem.
93
of more than 30 years. A similar argument is developed for the interaction between infrastructure for cars and the emergence of urban sprawl. The MARS model has been applied in various
studies to investigate optimal integrated transport and land-use strategies (see references within
Pfaffenbichler, Emberger, and Shepherd 2010) where impacts are assessed more rapidly than with
traditional models. It has also been linked with a fleet model by the same author called SERAPIS
so that choice of car as technologies change may also be studied (Pfaffenbichler, Krutak, and
Renner 2011).
Along similar lines, Haghani, Lee, and Byun (2003a) present a regional model based on an analysis of causal relationships and a feedback loop structure connecting a large number of physical,
socio-economic, and policy variables. The model is made up of seven sub-models: population,
migration, households, employment, residential and commercial development, travel demand,
and traffic supply or congestion effects. The model was programmed in the DYNAMO simulation
language, and based on data from Montgomery county, MD. The model is well documented and
includes calibration analysis using historic data from 1970 to 1980 validated against 19801990.
Haghani, Lee, and Byun (2003b) then report robustness and sensitivity tests covering inputs such
as birth rate and regional economic growth rates. The model was used to assess the impacts of
highway capacity expansion and the related changes in land use which in turn affected demand
and the performance of the transport network.
Shen et al. (2009) develop a high-level model comparing low/high density land use policies
for Hong Kong. In the distant long term they show that compact high density scenarios are more
sustainable with investment in rail-based transport over car infrastructure which is seen to be
more prominent in the low density case. The growth in population in the very long term can be
accommodated only in the high density scenario so the authors conclude: that only by means of
a planning policy scheme to support compact and high-density development could Hong Kong
meet the environmental, social and economical requirements of sustainable land use and achieve
a perfect balance among them.
Wang, Lu, and Peng (2008) again develop a very high-level interaction model between population, vehicle ownership, environment, GDP, travel demand and infrastructure supply, applying it
to a case study of Dalian, China. Car ownership policies are studied and the wider system effects
mean that strict ownership controls results in a larger city with a greater gross domestic product
(GDP) and greater share of public transport. It is noticeable that by 2050 the interplay between
population, car ownership, congestion, pollution and GDP mean that there are a similar number of
vehicles in total but for quite different populations. Cities with vehicle ownership encouraged have
a population of around 3 million versus 9 million inhabitants where ownership is not encouraged,
however it appears that the larger city has a lower GDP per capita.
Feng and Hsieh (2009) propose a hybrid model which integrates SD, cognitive maps, and
a sensitivity model to study the problem of investment in transport systems and in particular the question of resource allocation over time or when to invest. A case study for Taipei is
used to illustrate the approach to satisfying the needs of multiple stakeholders. The results of
the sensitivity analysis revealed that the increase in private vehicle trips caused increases in
emissions, energy consumed and accidents and so failed to meet the competing needs of multiple stakeholders. However, the system is shown to be insensitive to managers decisions on
resource allocation timings i.e. delays to implementation are not the problem. Policies which
control the growth in car use are shown to be the most effective at meeting the needs of
stakeholders.
All the above studies deal with short-term policies which have long-term impacts on urban
form or infrastructure and vice versa. System dynamics is ideal for investigating such systems
which contain feedbacks and delays which are often outside of the mental model of the decision
maker or where feedbacks cross stakeholder boundaries. Pataki et al. (2009) take the interaction
94
S.P. Shepherd
one step further and include the impact of forests. A multi-disciplinary study which used mediated
model building over a period of six months to build a model of the transport and ecosystem of
fossil fuel emissions is reported. The team conducted a whole ecosystem study of the role of
climate, urban expansion, urban form, transportation, and the urban forest in influencing net CO2
emissions in the Salt Lake Valley, Utah. The study showed the importance of the positive feedback
between urban developments and investment in transport infrastructure. Emissions were seen to
increase as this feedback created higher densities and increased traffic. The results suggested that
a strategy of doubling the density of tree planting would not have a significant impact on total
urban CO2 emissions, while land use and transport policies which combine to reduce urban sprawl
could produce a 22% reduction in CO2 emissions by 2030 compared with the business as usual
scenario.
Continuing with the theme of policies aimed at reducing environmental impacts, Han et al.
(2008) investigate emission reduction policies for the Inter-city transport problem. Sensitivity
analysis showed that reduction in CO2 is best achieved by increased development of the railway
network, but with a reduced emphasis on highway extensions and by imposing increases in fuel
taxes. Piattelli et al. (2002) study carbon taxes in the context of freight in Germany. Their model
covers investment in infrastructure, road, rail and waterway, subsidies, fuel costs and carbon taxes.
Carbon tax is shown to have only small impacts on mode share and raising tax on all fuels actually
increases the road share due to the relative costs of fuel by mode giving the counterintuitive result
that a negative carbon tax for freight movements would reduce road share. Carbon tax simply
increases revenues raised which could be used to reduce CO2 by other means but the study suggests
that investment in infrastructure for rail is better than a carbon tax, but even this has only small
impacts on mode shares.
Egilmez and Tatari (2012) also aim to reduce CO2 (in the USA), and three potential strategies
are tested with different levels of intervention: fuel efficiency, public transportation and electric
vehicle usage. Figure 8 gives an overview of the stock-flow diagram for their model showing
the links between population, GDP, wealth, vehicle stock, vehicle miles travelled, congestion,
emissions, highway capacity, and land use. Population and GDP growth rates are exogenous
while other equations are given in detail in this paper. The results showed that to be successful,
some hybrid of individual policies was crucial. All three policy approaches are required to meet
95
the CO2 targets in the USA 2 alone is not enough three are needed with at least two of them at
the 2nd level of implementation/success i.e. most optimistic. The authors also recognise the need
to change the objective from a traditional CBA approach to one which accounts for sustainability
if policies are to be adopted. This links nicely with a more general SD model of policy resistance
to change, Harich (2010); which discusses how social forces which favour change are inter-linked
with those which favour resistance to change and that a proper coupling is required to understand
why certain policies are ineffective.
European wide models have also been built to investigate EU wide transport and economic
developments. Fiorello, Fermi, and Bielanska (2010) provide an overview of the ASTRA model.
ASTRA links transport demand, the economy, with the vehicle fleet and environmental impacts
calculated for the European level. ASTRA has been used over the past decade to inform the
European Commission about the impacts of European transport policy especially over a long
time horizon, providing evidence for the White Paper on Transport. It is often used in conjunction
with other models also developed using a SD approach, such as the world energy model POLES
and the fleet development model TREMOVE. It is encouraging to note that whilst traditional
models such as the four-stage transport model still dominate the market in terms of research and
application areas that clients such as the EU are taking note of other approaches such as these
based on a SD perspective.
Finally within this section the policy of road pricing is studied by Shepherd (2013) who firstly
shows that a standard goal seeking structure from SD is equivalent to transport dis-equilibrium
modelling approaches in developing a simple model of competing toll operators. Shepherd then
shows how estimating response elasticities during periods of change or prior to a full equilibrium
can lead to errors in the toll setting strategies and cause instability and sub-optimal toll levels.
Liu, Triantis, and Sarangi (2010) develop a non-standard model of congestion pricing, combining
linguistic and fuzzy preferences plus social networking effects to model mode choice. Figure 9
shows part of the model covering peoples perception impacting on mode choice and the flow
Figure 9. Effect of peoples perceptions on the choice between bus and car.
Source: Liu, Triantis, and Sarangi (2010).
96
S.P. Shepherd
of revenues for reinvestment of toll revenue into the bus and Metro system which is modelled
over time. This demonstrates how SD can be used with other modelling approaches to provide a
different angle or viewpoint for a common policy issue.
97
while the trend extrapolation approach overshoots the peak; and the regression model misses
it completely. Compared with Liehr et al., Lyneis model includes other reinforcing loops such
as competing airlines all attempting to gain market share during the upturns which then results
in overall capacity being overshot; the competition for new aircraft can lead to a longer lead
time which then leads to the need to project further ahead which all combine to amplify the
cycles.
Agusdinata and de Klein (2002) develop a causal loop analysis of airline alliances covering
the rush for globalisation during a transition economy, the impacts of congestion at hub airports,
uncertainty and distrust and the potential for collaborative learning. The analysis is linked to the
main literature in the air industry and the paper demonstrates how a CLD can be developed to
explain complex behaviour within a cyclic market. The paper whilst useful for discussing market
structure and responses, leaves the quantification of the model to future research.
Pierson and Sterman (2013) develop a model of airline industry profit cycles since de-regulation
in the US. The model expands the boundaries of previous work and contains four main elements,
endogenous capacity, demand, pricing and costs. The paper details the parameter estimation
process and the use of Markov Chain Monte Carlo methods to establish confidence intervals.
Contrary to prior work they show that the delay in aircraft acquisition is not as influential a
determinant of the profit cycle as others have found and that price setting strategies were found to
play a surprisingly important role in stabilising profits. These papers demonstrate the advantages
of SD approaches in that they are able to capture the cyclical market within the structure of the
model which provides more useful insights into strategies to deal with cycles as shown in Pierson
and Sterman (2013).
Investment in airport runway capacity has been studied with a simple model by Miller and
Clarke (2005, 2007). While both papers use the same model, the better description is in the later
paper. This paper develops and illustrates via a simple model a methodology for assessing the
strategic value of air transportation infrastructure, in particular the benefits of being able to react
swiftly to changes in the market. The model includes the influence of airport capacity on airline
98
S.P. Shepherd
congestion costs which in turn affect the passenger fares and level of service, see Figure 11. These
factors then determine the aircraft per hour which in turn affect airport revenues. The model is
used to assess different strategies for infrastructure delivery. These strategies include: the level
of capacity increase, the time required to implement the capacity and the congestion threshold
which triggers the need for additional capacity. Monte Carlo simulation is used to account for
multiple sources of uncertainty. The model showed that a strategy of capacity enhancement based
on small increments and shorter response times could yield greater benefits than strategies that
consider larger capacity increases and which require longer response times. Future work could be
to link to an air market model. The earlier paper (Miller and Clarke 2005) combines real option
analysis via Monte-Carlo simulation with the model. The simulation is used to evaluate when the
option of building a second runway is greater than the real option of buying the land. In general
it shows that when an inflexible investment has an net present value (NPV) close to zero then the
flexibility to delay or advance a project can be most valuable. The issue of when to implement
large projects is something which is common in transport problems and SD models are a useful
tool in this area.
Suryani, Chou, and Chen (2010) extend this model to include impacts of population growth
and GDP and the impacts on both runway and terminal capacity. The model is used in a scenario
planning mode with optimistic and pessimistic cases. The difference in capacity requirements
suggested that the runway may cope with demand for an additional eight years in the pessimistic
case, and factors such as level of service requirements can have a significant impact on the floor
space required for the terminal. Suryani, Chou, and Chen (2012) adapt their model to investigate
air cargo demand and include effects of competition with other airports. This paper includes
a validation of the model against historic data along with a scenario analysis with high/low
growth scenarios to investigate when to invest in additional terminal capacity. A comparison with
statistical extrapolation approaches suggests the low growth scenario is more probable but the
authors suggest that the SD model is more useful as it includes the underlying structure of the
problem and hence the possibility to understand the implications of a more optimistic scenario
and the factors which influence the need for additional capacity.
Manataki and Zografos (2009, 2010) report on a generic planning tool for airport terminal
analysis. Whilst applied to a case study of Athens airport, the tool is designed to be generic
and easily applied to other airport terminal layouts. Their aim was to bridge the gap between a
very detailed operational analysis and a too aggregate macroscopic approach. The model splits
the terminal into a series of functional areas and facilities and is able to capture interactions
between these processes whilst allowing for random arrivals, delays, variations in schedule,
capacity and levels of service as well as basic design parameters. The approach uses a flight simulator or GUI to facilitate what if scenarios for the planner/decision-makers. Finally, Sgouridis,
Bonnefoy, and Hansman (2011) present the Global aviation industry dynamics (GAID) model
which includes three sectors, passengers, airlines and manufacturers. The study examines five
generic policies which aim to reduce the emissions of commercial aviation. Strategies investigated include, technological efficiency improvements, operational efficiency improvements, use
of alternative fuels, demand shift and carbon pricing or market-based incentives. Each policy
was implemented with three levels of effectiveness and combinations were tested and impacts
on CO2 , revenue passenger-kms, profitability and average fares were recorded. An interesting
aspect of the dynamics was the response to the highest level of carbon pricing where airline
profits suffer initially, but as airlines shed aircraft in response they are then able to set higher
fares and profits return to previous levels. This corresponds to real-world behaviour seen when
fuel prices were seen to double in 2008, initial disruption and some bankruptcies, followed
by mergers which effectively reduced capacities were observed. It seems that due to the long
timescales involved in the airline industry and in airport infrastructure coupled with changing
99
demands that SD has a lot to offer in this area and that more effort should be directed at such
issues.
8. Miscellaneous/emerging areas
As transportation covers a wide area, it has not been possible to categorise some papers within
the themes or areas identified above. This section includes some other application areas which
are very specific or which may point the way to new emerging areas for the application of SD in
the transportation field.
At the more detailed end of the transportation spectrum, Mehmood, Saccomanno, and Hellinga
(2003) apply SD to the area of car-following models. The proposed model assumed that drivers
were capable of estimating the spacing between their own vehicle and the next downstream vehicle
and rather than being based on drivers avoiding a collision, it is based on perception of safety. It
does not include the lead vehicles speed but does include status of brake lights. The model was
calibrated and validated against field data and proved to work well for three vehicles. However it is
not clear whether it would be practical to build a full micro-simulation model using this approach
and a more standard micro-simulation approach which takes on board the model suggested may
prove more useful in future work.
He et al. (2011) step up a level to investigate the problem of pollution around toll plazas in
China. The authors combine a computational fluid dynamics model with a SD model of toll plaza
operation to investigate the pollutants emitted under different assumptions about the penetration
rates and capacity (number of lanes) for electronic toll collection technologies. The SD part
was used to represent the queuing and service delays as vehicles go through the toll plaza. It
was demonstrated that there was some interaction between penetration rate and number of lanes
assigned to electronic toll collection which meant that pollutants may increase around the toll
booths as more lanes are provided under certain penetration rates.
Changing mode, Macmillan et al. (2014) used participatory group modelling to from a qualitative model of the major causal loops in the take up of cycling in Auckland, New Zealand. The loops
included a safety in numbers effect which reduces the real and perceived injury rate as cyclist
numbers increase. The model was populated with best available data and calibrated to recent
historic trends. Policies to encourage cycling were modelled over a 40-year period demonstrating
impacts on health, fuel savings and carbon emissions. It was meant to provide policy insight
rather than falsely precise predictions and sensitivity tests were used to check for any parameters
which either changed the ranking of policies or the order of magnitude of the outcomes. In this
way the model was useful for exploring policy and model assumptions and provided a basis for
discussions around cycling with others cities such as London.
Goh and Love (2012) develop two models to investigate policies to improve traffic safety.
The first model is used to assess policy options which aim to encourage the purchase of cars
with higher safety ratings. The second model is used to evaluate the impact of public transport
policies on travel time and traffic safety considerations. The first model looks at how cars with
different safety ratings are purchased and is similar to the work on alternative fuel uptake models,
network effects and price effects being included. The larger the current share the more attractive
it is to purchase due to lower maintenance costs and social norm factors. It was shown that
high taxes and subsidies for low-/high-rated cars are needed to overcome the network and price
effect. However, the sensitivity tests suggested that policies aimed at reducing the network effect,
e.g. awareness or early scrappage schemes for low-rated vehicles may improve the fleet average
star rating more quickly. The second model combined a goal-based safety policy which reacts
to a number of crashes versus a tolerable level but then added the link between this and the
policy on subsidies for public transport. It was shown that the larger the subsidy the greater the
100
S.P. Shepherd
reduction in crashes and hence less need for other safety interventions which could be more
expensive per saved crash. Similar to the work on the take up of AFVs, this paper demonstrates
the importance of long-term feedbacks and of taking a more holistic approach to the problem of
safety.
Bivona and Montemaggiore (2010) develop a case-study analysis of maintenance policies for
a city bus company. They describe the links between fleet maintenance, human resources and
training as well as other key policies such as fleet renewal and how these may impact on service
provision, customer satisfaction and overall profitability of the company. Two scenarios were
developed to try and prevent the decline in revenues. The first was to cut all departmental budgets
which while being effective in the short term were shown to be unsustainable in the long term.
This is a classic use of SD where performance measures are seen to get worse before then getting
better. The second policy included some investment in new buses, which reduced the average age
of the fleet, along with investment in training, all of which allowed increased service levels and
was shown to be more sustainable. This paper showed how the SD approach could be used with
decision makers to help them move away from simple cost-cutting policies to more sustainable
policies which at first sight appeared to be counter-intuitive.
Mayo, Callaghan, and Dalton (2001) report on a model for London Underground Limited
who understood that to evaluate different restructuring options would require a holistic and integrated system-wide view of the underground and its stakeholders. The model covered a range of
stakeholders including the market or customers who had a choice of mode, a workforce module including morale, experience, etc. suppliers, government and other private funding/finance.
Aspects such as capital assets and maintenance played an important role. Profit motivation and
investment levels could go into either virtuous or vicious circles depending on relative strengths
of feedback loops. The model was used to look at re-structuring options for LUL. Asset management, staffing continuity and collaboration early on were seen to be crucial in the success
of implementation of a public private partnership and hard to reverse. The model was provided
to the bidders to improve the tender process. The basic message to bidders from the model was
that an infrastructure company that maintains and invests appropriately in the asset base; maintains staff continuity; partners collaboratively with LUL and meets or exceeds the LUL-specified
performance targets will achieve significantly greater profits than one who does not. This paper
provides a good example of a model used in practice.
Randers and Gluke (2007) in a similar vein as the air market cycles, explain shipping markets
since 1950 as the interaction between two balancing feedback loops: a capacity adjustment loop
which creates a roughly 20-year wave, and a capacity utilisation adjustment loop which generates
a roughly 4-year cycle. They describe how they persisted with a small model rather than use a
detailed model and provided forecasts of when freight rates peaked. Shipping clients were at first
distrusting of the model but senior executives were convinced by the simpler models and some of
the client group trusted the forecasts to decide when to invest and when to get out of the market.
Another example of real-life impact and insight from a simple model which takes into account
stocks and flows.
Finally, Yeo, Pak, and Yang (2013) develop a model of the impacts of port security on the
competitiveness of the port. A CLD is developed with stakeholders and the quantitative model is
used to explore the impacts of increased security which on the one hand increases attractiveness
but on the other reduces it due to increased processing time and costs. Applied to a case study in
Korea, the study suggests that a more optimistic approach to security would increase trade and
prevent a spiral of decreasing competitiveness. Again the issue of feedback and policy resistance
is key to understanding the issues and in explaining how a more controversial or less favoured
policy may have the opposite impact to that expected by decision-makers whose mental model is
unable to take these factors into account.
101
10. Summary
In the 20 years since Abbas and Bell outlines the possibilities for the application of SD in transportation modelling, there has been a growing literature. Papers have appeared in some areas
more than others, as seen in the recent surge in the modelling of uptake of AFVs. This is in part
driven by the policy interest but is also an area which is well suited to the SD approach as it makes
use of existing structures such as the Bass diffusion model which has been applied to adoption
of new technologies in other areas. This structure is easily linked with choice models and fleet
ageing chains within the SD approach thus naturally accounting for time lags within the system.
As predicted by Abbas and Bell, strategic policy issues at a regional or national level involving
delays and feedbacks between different systems such as land use and transport have also been
developed as an area where SD has something to offer. Other areas which are well known for
their dynamics such as business cycles in the air market or for delays in responses as in the supply
chain area have also developed clusters of papers. The above review has outlined how the use of
both more qualitative CLDs and quantitative stock-flow models have been used to provide new
insights and explain the underlying structure within such systems. The approach often involves
stakeholders in the development of the CLDs and the simple nature of the stock-flow approach
aims at providing a transparent approach to modelling which is highly valued by stakeholders.
In the miscellaneous or emerging area section, it was seen that some applications were not
perhaps best tackled via SD. For example, SD is not appropriate for the traditional network
assignment problems, nor will it be able to replace micro-simulation tools. Whilst models can be
formed with spatial elements by making use of subscripts or arrays, from experience it seems that
once the number of zones increases beyond 200 then the model run times become a barrier and
large models are then run in compiled mode. Thus while large models are feasible in this mode,
it seems to go against the notion of a white box model and loses the benefits of communication to
102
S.P. Shepherd
stakeholders which comes with the use of smaller faster models which demonstrate the dynamic
behaviour of a system structure.
Instead the approach is better suited to providing a holistic system model which deals with
feedbacks and delays between actors in the system. Future applications could look at competition
dynamics, freight and the development of ports, sensitivity of systems and transport demand to
changing external factors related to demographics and the economy and in modelling behavioural
change whether this is at the user level of some higher level stakeholder. The approach allows
transport models to be easily linked to other sectors such as health, climate, and the economy
while taking into account time delays and feedbacks at different scales. However, it should be
noted that SD is not meant for precise point forecasts as Sterman (2002) summarises, System
dynamics helps us expand the boundaries of our mental models so that we become aware of and
take responsibility for the feedbacks created by our decisions.
System dynamics should be used to understand and explore the nature of the problem and gives
the modeller the opportunity to investigate general dynamic tendencies. The models can be used
to test which parameters play a significant role in the stability and response of the system and
the tools such as CLD and stock-flow diagrams enable a transparent approach to communicating
results with stakeholders including the use of flight simulators and gaming tools which other
approaches often lack. It is hoped that more transport modellers will in future work with the SD
approach and fulfil the aim of the international dynamics society which is to have a real-world
impact on business and policy. Finally, it is encouraging that clients such as the EU commission
and local authorities such as Leeds in the UK are commissioning SD based models to investigate
their specific problems, having seen the advantages of communication with their local elected
representatives.
Notes
1.
Rather than detail the various software platforms available the International system dynamics society
provides a list of key products at the following link: http://tools.systemdynamics.org/core-sd-software/
2. Bullwhip effect refers to a trend of larger and larger swings in inventory in response to changes in
customer demand, as one looks at firms further back in the supply chain for a product.
References
Abbas, K. A., and M. G. H. Bell. 1994. System Dynamics Applicability to Transportation Modeling.
Transportation Research Part A 28 (5): 373390.
Agusdinata, B., and W. de Klein. 2002. The Dynamics of Airline Alliances. Journal of Air Transport
Management 8 (4): 201211.
Barlas, Y. 1996. Formal Aspects of Model Validity and Validation in System Dynamics. System Dynamics
Review 12 (3): 183210.
Bivona, E., and G. B. Montemaggiore. 2010. Understanding Short- and Long-Term Implications of
Myopic Fleet Maintenance Policies: A System Dynamics Application to a City Bus Company.
System Dynamics Review 26 (3): 195215.
Chasey, A. D., J. M. de La Garza, and D. R. Drew. 2002. Using Simulation to Understand the Impact of
Deferred Maintenance. Computer-Aided Civil and Infrastructure Engineering 17 (4): 269279.
Disney, S. M., A. T. Potter, and B. M. Gardner. 2003. The Impact of Vendor Managed Inventory on Transport
Operations. Transportation Research Part E 39 (5): 363380.
Egilmez, G., and O. Tatari. 2012. A Dynamic Modeling Approach to Highway Sustainability: Strategies to
Reduce Overall Impact. Transportation Research Part A 46 (7): 10861096.
Fallah-Fini, S., H. Rahmandad, K. Triantis, and J. M. de la Garza. 2010. Optimizing Highway Maintenance
Operations: Dynamic Considerations. System Dynamics Review 26 (3): 216238.
Feng, C. M., and C. H. Hsieh. 2009. Effect of Resource Allocation Policies on Urban Transport Diversity.
Computer-Aided Civil and Infrastructure Engineering 24 (7): 525533.
103
Fiorello, D., F. Fermi, and D. Bielanska. 2010. The ASTRA Model for Strategic Assessment of Transport
Policies. System Dynamics Review 26 (3): 283290.
Flynn, P. 2002. Commercializing An Alternate Vehicle Fuel: Lessons Learned From Natural Gas for
Vehicles. Energy Policy 30 (7): 613619.
Forrester, J. 1958. Industrial Dynamics A Major Breakthrough for Decision Makers. Harvard Business
Review 35 (4): 3766.
Friedman, S. 2006. Is Counter-Productive Policy Creating Serious Consequences? The Case of Highway
Maintenance. System Dynamics Review 22 (4): 371394.
Georgiadis, P., D. Vlachos, and E. Iakovou. 2005. A System Dynamics Modeling Framework for
the Strategic Supply Chain Management of Food Chains. Journal of Food Engineering 70 (3):
351364.
Ghaffarzadegan, N., J. Lyneis, and G. P. Richardson. 2011. How Small System Dynamics Models Can Help
the Public Policy Process. System Dynamic Review 27 (1): 2244.
Goh, Y. M., and P. E. D. Love. 2012. Methodological Application of System Dynamics for Evaluating
Traffic Safety Policy. Safety Science 50 (7): 15941605.
Haghani, A., S. Y. Lee, and J. H. Byun. 2003a. A System Dynamics Approach to Land Use / Transportation
System Performance Modeling. Part 1: Methodology. Journal of Advanced Transportation 37 (I):
141.
Haghani, A., S. Y. Lee, and J. H. Byun. 2003b. A System Dynamics Approach to Land Use / Transportation
System Performance Modeling. Part 2: Application. Journal of Advanced Transportation 37 (I): 4382.
Han, J., and Y. Hayashi. 2008. A System Dynamics Model of CO2 Mitigation in Chinas Inter-City
Passenger Transport. Transportation Research Part D 13 (5): 298305.
Hang, W., and X. Li. 2011. Application of System Dynamics for Evaluating Truck Weight Regulations.
Transport Policy 17 (4): 240250.
Harich, J. 2010. Change Resistance as the Crux of the Environmental Sustainability Problem. System
Dynamics Review 26 (1): 3572.
Harrison, G., and S. P. Shepherd. 2014. An Interdisciplinary Study to Explore Impacts From Policies for
the Introduction of Low Carbon Vehicles. Transportation Planning and Technology 37 (1): 98117.
He, J., Z. Qi, W. Hang, M. King, and C. Zhao. 2011. Numerical Evaluation of Pollutant Dispersion at a
Toll Plaza Based on System Dynamics and Computational Fluid Dynamics Models. Transportation
Research Part C 19 (3): 510520.
Janssen, A., S. F. Lienin, F. Gassmann, and A. Wokaun. 2006. Model Aided Policy Development for the
Market Penetration of Natural Gas Vehicles in Switzerland. Transportation Research Part A: Policy
and Practise 40 (4): 316333.
Kohler, J., M. Wietschel, L. Whitmarsh, and W. Schade. 2010. Infrastructure Investment for a Transition to
Hydrogen Auotomobiles. Technological Forecasting & Social Change 77 (8): 12371248.
Kwon, T. 2012. Strategic Niche Management of Alternative Fuel Vehicles: A System Dynamics Model of
the Policy Effect. Technological Forecasting & Social Change 79 (9): 16721680.
Liehr, M., A. Groler, M. Kleinb, and P. M. Millinga. 2001. Cycles in the Sky: Understanding and Managing
Business Cycles in the Airline Market. System Dynamics Review 17 (4): 311332.
Liu, S., K. P. Triantis, and S. Sarangi. 2010. A Framework for Evaluating the Dynamic Impacts of a
Congestion Pricing Policy for a Transportation Socioeconomic System. Transportation Research Part
A 44 (8): 596608.
Lyneis, J. M. 2000. System Dynamics for Market Forecasting and Structural Analysis. System Dynamics
Review 16 (1): 325.
Macmillan, A., J. Connor, K. Witten, A. Kearns, D. Rees, and A. Woodward. 2014. The Societal Costs and
Benefits of Commuter Bicycling: Simulating the Effects of Specific Policies Using System Dynamics
Modeling. Environmental Health Perspectives 122 (4): 335344.
Manataki, I. E., and K. G. Zografos. 2009. A Generic System Dynamics Based Tool for Airport Terminal
Performance Analysis. Transportation Research Part C 17 (4): 428443.
Manataki, I. E., and K. G. Zografos. 2010. Assessing Airport Terminal Performance Using a System
Dynamics Model. Journal of Air Transport Management 16 (2): 8693.
Mayo, D. D., M. J. Callaghan, and W. J. Dalton. 2001. Aiming for Restructuring Success at London
Underground. System Dynamics Review 17 (3): 261289.
Mehmood, A., F. Saccomanno, and B. Hellinga. 2003. Application of System Dynamics in Car-Following
Models. Journal of Transportation Engineering 129 (6): 625634.
Meyer, P. E., and J. Winebrake. 2009. Modeling Technology Diffusion of Complementary Goods: The Case
of Hydrogen Vehicles and Refueling Infrastructure. Technovation 29 (2): 7791.
104
S.P. Shepherd
Miller, B., and J. P. Clarke. 2005. Investments under Uncertainty in Air Transportation: A Real Options
Perspective. Journal of the Transportation Research Forum 44 (1): 6174.
Miller, B., and J. P. Clarke. 2007. The Hidden Value of Air Transportation Infrastructure. Technological
Forecasting & Social Change 74 (1): 1835.
Otto, A., and H. Kotzab. 2003. Does Supply Chain Management Really Pay? Six Perspectives to Measure
the Performance of Managing a Supply Chain. European Journal of Operational Research 144 (2):
306320.
Park, S. Y., J. W. Kim, and D. H. Lee. 2011. Development of a Market Penetration Forecasting Model for
Hydrogen Fuel Cell Vehicles Considering Infrastructure and Cost Reduction Effects. Energy Policy
39 (6): 33073315.
Pataki, D. E., P. C. Emmi, C. B. Forster, J. I. Mills, E. R. Pardyjak, T. R. Peterson, J. D. Thompson, and E.
Dudley-Murphy. 2009. An Integrated Approach to Improving Fossil Fuel Emissions Scenarios with
Urban Ecosystem Studies. Ecological Complexity 6 (1): 114.
Pfaffenbichler, P. 2011. Modelling with Systems Dynamics as a Method to Bridge the Gap between Politics,
Planning and Science? Lessons Learnt from the Development of the Land Use and Transport Model
MARS. Transport Reviews: A Transnational Trans-disciplinary Journal 31 (2): 267289.
Pfaffenbichler, P., G. Emberger, and S. P. Shepherd. 2010. A System Dynamics Approach to Land Use
Transport Interaction Modelling: The Strategic Model MARS and Its Application. System Dynamics
Review 26 (3): 262282.
Pfaffenbichler, P., R. Krutak, and S. Renner. 2011. Modelling the Development of Vehicle Fleets with
Alternative Propulsion Technologies. ECEEE 2011 summer study, Belambra Presqule de Giens,
France.
Piattelli, M. L., M. A. Cuneo, N. P. Bianchi, and G. Soncin. 2002. The Control of Goods Transportation
Growth by Modal Share Re-planning: The Role of a Carbon Tax. System Dynamics Review 18 (1):
4769.
Pierson, K., and J. D. Sterman. 2013. Cyclical Dynamics of Airline Industry Earnings. System Dynamics
Review 29 (3): 129156.
Potter, A., and C. Lalwani. 2008. Investigating the Impact of Demand Amplification on Freight Transport.
Transportation Research Part E 44 (5): 835846.
Randers, J., and U. Gluke. 2007. Forecasting Turning Points in Shipping Freight Rates: Lessons From 30
Years of Practical Effort. System Dynamics Review 23 (2/3): 253284.
Sgouridis, S., P. A. Bonnefoy, and R. J. Hansman. 2011. Air Transportation in a Carbon Constrained World:
Long-Term Dynamics of Policies and Strategies for Mitigating the Carbon Footprint of Commercial
Aviation. Transportation Research Part A 45 (10): 10771091.
Shen, Q., Q. Chen, B. Tang, S. Yeung, Y. Hu, and G. Cheung. 2009. A System Dynamics Model for the
Sustainable Land Use Planning and Development. Habitat International 33 (1): 1525.
Shepherd, S. P. 2013. Toll Competition and Dynamic Toll Setting Strategies. International Journal of
Sustainable Transportation 7 (3): 186203.
Shepherd, S. P., P. W. Bonsall, and G. Harrison. 2012. FactorsAffecting Future Demand for ElectricVehicles:
A Model Based Study. Transport Policy 20 (March): 6274. doi:10.1016/j.tranpol.2011.12.006
Stepp, M. D., J. J. Winebrake, J. S. Hawker, and S. J. Skerlos. 2009. Greenhouse Gas Mitigation Policies
and the Transportation Sector: The Role of Feedback Effects on Policy Effectiveness. Energy Policy
37 (7): 27742787.
Sterman, J. D. 1984. Appropriate Summary Statistics for Evaluating the Historical Fit of System Dynamics
Models. Dynamica 10 (2): 5166.
Sterman, J. 2000. Business Dynamics: Systems Thinking and Modeling for a Complex World. Boston:
Irwin/McGraw-Hill.
Sterman, J. D. 2002. All Models are Wrong: Reflections on Becoming a Systems Scientist. System
Dynamics Review 18 (4): 501531.
Struben, J., and J. D. Sterman. 2008. Transition Challenges for Alternative Fuel Vehicle and Transportation
Systems. Environment and Planning B-Planning & Design 35 (6): 10701097. www.envplan.com
Suryani, E., S. Y. Chou, and C. H. Chen. 2010. Air Passenger Demand Forecasting and Passenger Terminal Capacity Expansion: A System Dynamics Framework. Expert Systems with Applications 37 (3):
23242339.
Suryani, E., S. Y. Chou, and C. H. Chen. 2012. Dynamic Simulation Model of Air Cargo Demand Forecast
and Terminal Capacity Planning. Simulation Modelling Practice and Theory 28 (November): 2741.
Tako, A., and S. Robinson. 2012. The Application of Discrete Event Simulation and System Dynamics in
the Logistics and Supply Chain Context. Decision Support Systems 52 (4): 802815.
105
Trappey, A. J. C., C. Trappey, C. T. Hsiao, J. J. R. Ou, S. J. Li, and K. W. P. Chen. 2012. An Evaluation Model
for Low Carbon Island Policy: The Case of Taiwans Green Transportation Policy. Energy Policy 45
(June): 510515.
Ulli-Beer, S., F. Gassmann, M. Bosshardt, and A. Wokaun. 2010. Generic Structure to Simulate Acceptance
Dynamics. System Dynamics Review 26 (2): 89116.
Walther, G., J. Wansart, K. Kieckhafer, E. Schnieder, and T. S. Spengler. 2010. Impact Assessment in
the Automotive Industry Mandatory Market Introduction of Alternative Powertrain Technologies.
System Dynamics Review 26 (3): 239261.
Wang, J., H. Lu, and H. Peng. 2008. System Dynamics Model of Urban Transportation System and Its
Application. Journal of Transportation Systems Engineering and Information Technology 8 (3): 8389.
Wilson, M. C. 2007. The Impact of Transportation Disruptions on Supply Chain Performance. Transportation Research Part E 43 (4): 295320.
Wolstenholme, E. F. 2003. Towards the Definition and Use of a Core Set of Archetypal Structures in System
Dynamics. System Dynamics Review 19 (1): 726.
Xu, H., A. N. Mashayekhib, and K. Saeed. 1998. Effectiveness of Infrastructure Service Delivery Through
Earmarking: The Case of Highway Construction in China. System Dynamics Review 14 (2): 221255.
Yeo, G. T., J. Y. Pak, and Z. Yang. 2013. Analysis of Dynamic Effects on Seaports Adopting Port Security
Policy. Transportation Research Part A 49 (March): 285301.