Вы находитесь на странице: 1из 11

Available online at www.sciencedirect.

com
Available online
Available online at
at www.sciencedirect.com
www.sciencedirect.com

ScienceDirect
ScienceDirect
ScienceDirect
Transportation Research Procedia 00 (2016) 000–000
TransportationResearch
Transportation ResearchProcedia
Procedia00
21(2016)
(2017)000–000
216–226
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

2016 International Symposium of Transport Simulation (ISTS’16 Conference), June 23~25, 2016
2016 International Symposium of Transport Simulation (ISTS’16 Conference), June 23~25, 2016

Junction parameter calibration for mesoscopic simulation in Vissim


Junction parameter calibration for mesoscopic simulation in Vissim
Anett Ehlert aa, Arne Schneck aa, Nate Chanchareon bb
Anett Ehlert , Arne Schneck , Nate Chanchareon
a
PTV Group, Haid- und Neu Str. 15, 76131 Karlsruhe, Germany
b
PTVa PTV Group,
Group, Haid-
18 Sin Mingund Neu#06-35,
Lane, Str. 15,Midview
76131 Karlsruhe, Germany
City, Singapore 573960.
b
PTV Group, 18 Sin Ming Lane, #06-35, Midview City, Singapore 573960.

Abstract
Abstract
In this paper, we present a mesoscopic simulation method with a special focus on the calibration of the junction control
In this paper,
parameters. we presentsimulation
Mesoscopic a mesoscopic
methodssimulation
are popularmethod withfast
for their a special
run timesfocus on thetocalibration
compared microscopic ofsimulations
the junctionmethods.
control
parameters.there
However, Mesoscopic
are somesimulation methods are
simpler behavioral popular forintheir
assumptions fast run times
the mesoscopic compared which
simulation, to microscopic simulations
will be described methods.
in the paper
However,
along withthere
the are some
basic simpler behavioral
algorithms assumptions
of the mesoscopic in the mesoscopic
simulation. Hereby, simulation, which will
we also describe the be described
main in theinpaper
differences the
along with the of
implementations basic algorithms
microscopic andofmesoscopic
the mesoscopic
simulationsimulation. Hereby, we in
as being implemented also
the describe the main
traffic simulator differences
software in the
PTV Vissim.
implementations
The main focus of of this
microscopic
paper is and mesoscopic
directed towardssimulation
parameterascalibration
being implemented in theThe
for junctions. traffic simulator
influence software PTV
of parameter Vissim.
variation is
The main focus
demonstrated forof this paper
different typesis of
directed towards
intersections parameter
and ranks of calibration
movements,for junctions.
namely The influence
this concerns of parameter
the follow-up and variation
critical gapis
demonstrated
and the maximum for different types More
waiting time. of intersections andit ranks
specifically, can beof movements,
shown namely
that adjusting thisthree
these concerns the follow-up
parameters allows toand critical
closely gap
match
and the maximum
illustrative exhibitswaiting
betweentime. More specifically,
conflicting volumes and it can
thebe shown that
potential adjusting
capacity these
in the three parameters
Highway allows to
Capacity Manual closelyDue
(HCM). match
to
illustrative exhibits between conflicting volumes and the potential capacity in the Highway Capacity Manual
the simplifications compared to microscopic algorithms junction modelling in mesoscopic simulation is based on the same gap (HCM). Due to
the simplifications
parameters compared
as the analytic to microscopic
approach algorithms
in the HCM. junction modelling
Consequently, in mesoscopic
default values simulationinis Vissim
for these parameters based on the asame
form gap
suitable
parameters as the analytic
basis for parameter approach
calibration in situations.
to local the HCM. Consequently, default values for these parameters in Vissim form a suitable
basis for parameter calibration to local situations.
© 2016 The Authors. Published by Elsevier B. V.
Copyright
© 2016 The© 2017 The Authors.
Authors. Published
Published by by Elsevier
Elsevier B. V.B.V.
Selection
Selection and
and Peer-review
Peer-review under
under responsibility
responsibility of of
of Dept. Dept. of Transportation
Transportation Engineering,
Engineering, UniversityUniversity
of Seoul. of Seoul.
Selection and Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul.
Keywords: PTV Vissim; mesoscopic simulation; model calibration; junction performance
Keywords: PTV Vissim; mesoscopic simulation; model calibration; junction performance

1. Introduction
1. Introduction
In transportation modelling different approaches based on the level of detail are considered. On one side there
areIn transportation
microscopic modelling
models where different
individualapproaches based
vehicles and theironinteractions
the level ofwith
detail are neighbors
their considered.
areOn one side there
considered. The
are
core algorithms are car following models and lane changing algorithms that are used to move vehicles throughThe
microscopic models where individual vehicles and their interactions with their neighbors are considered. the
core algorithms
network. One the areother
car following models and
side macroscopic lane exist
models changing
where algorithms thatquantities
aggregated are used to move
such vehicles speed,
as density, throughflow
the
network. One the other side macroscopic models exist where aggregated quantities such as density,
and the relationships between them are used to assign vehicles to a network. For microscopic and macroscopic speed, flow
and the there
models relationships
has beenbetween
a commonthem are used toof
understanding assign vehicles to
the underlying a network. For microscopic and macroscopic
algorithms.
models there has been a common understanding of the underlying algorithms.

2214-241X © 2016 The Authors. Published by Elsevier B. V.


2214-241X © Peer-review
Selection and 2016 The Authors. Published by of
under responsibility Elsevier B. Transportation
Dept. of V. Engineering, University of Seoul.
Selection and Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul.

Copyright © 2017 The Authors. Published by Elsevier B.V.


Selection and Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul.
10.1016/j.trpro.2017.03.091
Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226 217
2 Anett Ehlert / Transportation Research Procedia 00 (2016) 000–000
The mesoscopic level that represents the middle layer between these approaches is less well defined as a variety
of approaches claim to represent an intermediate level overcoming some of the shortcomings in microscopic and
macroscopic models, respectively.

2. Approaches to mesoscopic modelling

Mesoscopic modelling covers a variety of approaches implemented in different software tools from macroscopic
planning tools up to microscopic simulation tools. The approaches have in common that they aim to overcome
weaknesses of macroscopic models, e.g. produce more detailed output taking time information into account, and
microscopic models, e.g. reduce run times at the cost of very detailed output of vehicle data.
Examples of mesoscopic approaches are:
 models where individual vehicles move through the network based on aggregated macroscopic quantities
derived from the fundamental diagram such as density or aggregated speed, e.g. Burghout et al. (2006) and
references therein
 models where groups of vehicles with similar characteristics move through the network, e.g. Leonard et al.
(1989)
 cell-based approaches in which the road network is divided into cells, and depending on how many vehicles
fit in such a cell they are further subdivided into microscopic models (one vehicle per cell) or mesoscopic
models (more than one vehicle per cell), e.g. variants of the cellular automata models of Nagel and
Schreckenberg (1992), Mahut (2001) and references therein
 models where individual vehicles move through the network but based on simplified car following and lane
changing models, e.g. Mahut (2001)

3. Microscopic and mesoscopic simulation in PTV Vissim

3.1. Car following

For microscopic simulation PTV Vissim uses the psycho-physical perception model developed by Wiedemann
(1974) that distinguishes between four different driving states, namely free flow, following, approaching, and braking.
The basic concept of this model is that the driver of a faster moving vehicle starts to decelerate as he reaches his
individual perception threshold while approaching a slower moving vehicle in front. Since he cannot exactly
determine the speed of that vehicle, his speed will fall below that vehicle’s speed until he starts to slightly accelerate
again after reaching another perception threshold. Fig. 1 depicts this approaching process and the different
perception thresholds that separate the four driving states (PTV Group, 2015).

Fig. 1: Car following logic according to Wiedemann (1974) with the following driving states:

1-free flow, 2-following, 3-approaching, 4-braking, 5-collision


218 Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226
Author name / Transportation Research Procedia 00 (2016) 000–000 3

The mesoscopic simulation uses a simplified car following model based on Mahut (2001). This model
distinguishes between two states only: Either the vehicle moves with its desired speed as long as there is no slower
vehicle in front, or the vehicle follows a slower vehicle in front and keeps a bumper to bumper time headway that
corresponds to the reaction time plus the time it takes to travel the standstill distance.
This can be expressed by the following equation

��������� ��� � �������� � � ��������� � ������� �� � ��������� � � �������


� � �

where
�������� – front end of the following vehicle
������ – front end of the leading vehicle
�������� – desired speed of the following vehicle
�������� – reaction time of the following vehicle
������� – effective vehicle length (vehicle length plus standstill distance) of the leading vehicle In

contrast to the microscopic car following model, acceleration and deceleration is not modelled.

3.2. Lane changing and lane selection

In microscopic simulation there are two types of lane changes: mandatory lane changes that are required in order
for the vehicle to stay on its route, and free lane changes because of slower vehicles in front. In both cases the vehicle
needs to find a suitable gap in the destination lane. Consequently, the execution of a lane changing maneuver depends
on the interactions with vehicles in the same and neighboring lanes and characteristics of these vehicles.
In mesoscopic simulation mandatory lane changes are considered through lane selection at the upstream node
of the link. A vehicle selects a lane that allows to follow its route. More specifically, if downstream a certain turn
can only be reached from one lane and the vehicle must make this turn it enters the link on the corresponding lane.
If several lanes can be used the selection is made based on macroscopic quantities of the corresponding lanes. Free
lane changes are not modelled explicitly, i.e. vehicles can overtake each other if they select different lanes of the
same link, but not if they select the same lane.

3.3. Time resolution for network loading

In microscopic simulation vehicle data is updated at each time step (e.g. 10 time steps per second). This includes
the vehicle positions enabling on the one hand the visualization of all vehicle movements over the course of the
simulation time. On the other hand the updated vehicle position data is also processed for evaluating the interactions
with other vehicles.
In contrast, in mesoscopic simulation the car following model is solved differently. Instead of obtaining the
vehicle’s position after a fixed time step (i.e. discretizing the model in time), the time is computed when the vehicle
reaches a fixed position, e.g. the next node (i.e. the model is discretized in space). This leads to a so-called event-
based simulation. Events are instantaneous occurrences when properties of vehicles or the system changes.
Examples are a vehicle entering a node or arriving at the downstream end of a link, or changes of signals. All these
events are scheduled and either processed or re-scheduled in the order they occur over the simulation time. As these
events occur less frequent than fixed time steps in microsimulation, the calculation effort during mesoscopic
simulation is reduced considerably. However, the underlying resolution can still be as detailed as the microscopic
simulation.

3.4. Parameters for modelling behavior at intersections

The above listed differences between microscopic and mesoscopic simulation have consequences on the
parameters that influence the behavior of vehicles at intersections. In VISsim gap acceptance for give way situations
is modeled using so-called conflict areas. Generally, conflicts potentially arise where vehicles share the same road
Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226 219

4 Anett Ehlert / Transportation Research Procedia 00 (2016) 000–000


space, or speaking of the network model where links of different directions overlap. One can distinguish between
different types of conflicts, namely crossing, merging and branching. The vehicles of conflicting traffic streams are
either in a give way situation, i.e. vehicles from one traffic stream have the right of way, or they are equally entitled
to enter the common road space, i.e. the vehicle that arrives first enters the shared road space first. Conflict areas
are defined for pairs of links whereby the actual conflicts arise on a lane-by-lane basis.
For microscopic simulation several parameters for conflict areas are provided to reproduce the behavior of
vehicles when passing through an intersection. These parameters allow for vehicle class specific definitions and
apply to specific types of conflicts. The default values consider that driving behavior of vehicles is replicated to a
very detailed level by the car following and lane changing models. Specifically, this refers to the constant adaptation
of acceleration and deceleration at each time step.
For mesoscopic simulation only the parameter for the critical gap value of conflict areas applies to model the
interactions between vehicles of conflicting movements at intersections in give way situations. For conflicts without
right of way definitions it is sufficient to ensure that vehicles keep the minimum headway to each other determined
by the reaction time and the effective vehicle length.
To evaluate the total delay that a vehicle is experiencing when passing through an intersection two further
parameters are required, namely the follow-up gap and the maximum waiting time. The follow-up gap is defined
for the turn, and the maximum waiting time is defined for the link of the corresponding minor street approach. The
three parameters together are responsible for determining junction performance in mesoscopic simulation. Their
definitions are as follows:
 The parameter meso critical gap defines the minimum time interval between two vehicles of the major traffic
stream (rear edge of leader to front edge of follower) that allows one vehicle from a lower ranked movement
to turn into the desired direction. Consequently, the critical gap must be defined for conflicting movements
where vehicles must give way to traffic streams of higher ranked movements.
 The parameter meso follow-up gap is the time headway between the departures of two consecutive vehicles from
the same approach. The follow-up gap has only an effect if its value is greater than the time headway defined
by the car following model between two vehicles. More specifically, this parameter must be greater than the
reaction time plus the time it takes to travel the distance of the leader’s vehicle length.
 The parameter meso maximum waiting time defines the maximum time for a vehicle of the lower ranked
movement before it forces its entry into the higher ranked traffic stream.

4. Parameter calibration

The calibration of junction parameters is first demonstrated for a specific give way situation, namely the left
turn movement from the major stream at an unsignalized intersection. The outcome of simulation runs with varying
parameters is compared to the results from the analytic approach defined in the HCM (Highway Capacity Manual,
2010).
Further recommendations of how to calibrate parameters for other movements are given. Special attention is
paid to situations where the simulation models provide more detailed information than assumed in the analytical
procedures of the HCM.

4.1. Example intersection data

The following sketch in Fig. 2 represents the example intersection. The major flow movement is in the eastbound
direction. The movement marked with the black arrow is a rank 2 movement, left turn from the major direction for
which gap parameters are to be calibrated. Vehicles of this movement must give way to the oncoming traffic as
well as to vehicles that turn right. Without loss of generality for this case the conflicting flow in the simulation
consists of vehicles moving eastbound only. The conflict between these two streams is shown by the overlapping
road space within the junction.
220 Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226

Author name / Transportation Research Procedia 00 (2016) 000–000 5

Fig. 2: Example network for parameter calibration of Rank 2 movement (Major flow – left turn)

Table 1 summarizes input data to the simulation model. For simplicity, only cars with similar characteristics are
used. A total of 180 simulation runs has been carried out to evaluate different combinations of parameter sets with
varying conflicting flows.

Table 1. Parameters of example network

Parameter Default Range Step size


Vehicle length 4m n/a n/a
HGV percentage 0 n/a n/a
desired speed 50km/h n/a n/a
Reaction time 1.2s n/a n/a
Standstill distance 2.0m n/a n/a
Max. waiting time 120s 90 – 180s 30s
Critical gap 3.5s 2 – 5s 1s
Follow-up gap 2.2s 1.5 – 3.5s 0.5s
Volume from west eastbound (major) n/a 3,000 n/a
Volume from east southbound (major LT) n/a 3,000 200 – 300 vehs

4.2. The effect of varying parameters

Different parameter values for the three parameters relevant for junction performance are tested to compare the
results against the outcome based on the HCM analytic procedure. The simulations are run varying one of the three
parameters only, namely the critical gap, the follow-up gap and the maximum waiting time. For the other two
parameters the default values are used, respectively. Furthermore, the simulations are repeated increasing the
demand flow of the major traffic stream in the eastbound direction for each run by a fixed amount up to a maximum
of 3,000 vehicles. The demand flow for the left turning movement is kept at its maximum. The values for the potential
capacity of the rank 2 movement are determined by the number of vehicles that are able to make the left turn during
Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226 221

6 Anett Ehlert / Transportation Research Procedia 00 (2016) 000–000


the simulation run. Fig. 3 displays the results based on varying value for the critical gap parameter.

Fig. 3: Varying the critical gap value

It can be concluded from these curves that the parameter of the critical gap has a major influence on the shape
of the curves, meaning that the potential capacity for medium flow levels of the major street flow changes drastically
depending on this parameter. For example, the values of the potential capacity vary by approximately 600 vehicles/h
at conflicting flows of about 1,000 vehicles. The black curve corresponds to the use of default parameters in Vissim.
Fig. 4 displays curves using varying parameter values for the follow-up gap. From these curves it can be concluded
that this parameter defines the intersection point with the y-axis, i.e. the potential capacity if no conflicting flow is
present. The potential capacities decrease in a similar way down to a value of approx. 60 vehicles/h when flow
levels at the major flow reach maximum levels. Again here the black curve corresponds to the use of default
parameter values in Vissim.

Fig. 4: Varying the follow-up gap value


222 Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226
Author name / Transportation Research Procedia 00 (2016) 000–000 7

Finally, the parameter for the maximum waiting time is varied. From Fig. 5 it can be seen that this parameter
has the smallest effect on the potential capacity as the curves are nearly of identical shape. Small variations can be
observed at very high flow levels of the conflicting traffic stream. The highest value of 85 vehicles/h for the potential
capacity is reached with a maximum waiting time of 60s. This value decreases to 55 vehicles/h for a maximum
waiting time of 180s.

Fig. 5: Varying the maximum waiting time value

From this exercise it can clearly be seen that the three parameters have a direct impact on the potential capacity.
The highest influence comes from the definition of the critical gap parameter that greatly determines the capacity
over a wide range of conflicting flow levels. The follow-up gap can be used to determine the capacity of the minor
stream in the absence of vehicles of the major traffic stream. Little impact comes from the maximum waiting time
as capacities vary little even at high flow levels.

4.3. Calibration against HCM

In the previous section the influence of parameter values on the potential capacity was shown for the left turning
movement of rank 2. In the absence of observed data a comparison with the outcomes from the HCM gives a good
indication of the quality of the calibration efforts.
Generally, the HCM defines an analytical procedure to calculate junction performance indicators. For
unsignalized intersections the steps involve the definition of conflicts and the determination of conflicting flow rates.
It is important to note that in the HCM the definitions of conflicting flow rates are based on movements, i.e. do not
consider the exact junction topology and distribution of vehicles across lanes. Furthermore, depending on the type
of intersection and rank there are fixed proportions of conflicting flows assumed. In the example network according
to the HCM 100% of vehicles of the straight movement and 100% of vehicles turning right define the conflicting
flow for the left turning vehicles for which the potential capacity is derived.
The same is valid for the simulation network. Conflicting flows are defined by 100% of vehicles from the opposite
direction turning straight and 100% of vehicles from the opposite direction turning right. In other words, the
conflicting flows in the simulation correspond to the definition in the HCM for this movement. The next step in the
HCM derives the potential capacity assuming base values for the critical gap and the follow-up gap. If applicable,
adjustment factors for HGV percentage, grade or intersection geometry apply.
The curve from the test runs using default parameter values in Vissim are compared to the HCM graph using
the base critical headway (exhibit 19-10) and base follow-up headway (exhibit 19-11) for a two-lane major street
Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226 223
8 Anett Ehlert / Transportation Research Procedia 00 (2016) 000–000

intersection. Adjustment factors are not used for obtaining the HCM curve in Fig. 6. This is in line with the
assumptions made for the example network and corresponding test runs.

Fig. 6: Comparison between HCM and Vissim curves for the major street left turn movement of rank 2 using default parameter values

Fig. 6 compares the two curves obtained from the HCM analytical procedure and the simulation runs using the
default parameters for the critical gap, the follow-up gap and the maximum waiting time. Small differences can be
observed at high conflicting flow rates. This can be attributed to the fact that in the simplified HCM junction model
vehicles are assumed to be of zero length, which leads to unrealistically large numbers of long enough gaps especially
for high conflicting flow rates. This is not the case in the simulation runs and consequently, the potential capacity
is slightly lower at higher conflicting flow rates. Also note that the default critical gap values in Vissim are smaller
than the base critical gaps from the HCM. For the movement under consideration in the example network the HCM
defines a critical gap value of 4.1s while the default parameter value for the critical gap in Vissim is 3.5s. Again,
this is due to the assumption in the HCM that vehicles have zero length.
Nevertheless, the test runs with varying parameter values have shown that the parameters provided are suitable
to obtain realistic capacities for traffic streams that must give way. Furthermore, the use of default parameter values
already results in a good match with the outcomes of the well-accepted HCM analytical procedure provided the
assumptions in the HCM are applicable to the movement under consideration.

4.4. Differences between the HCM and simulation

For the analytical procedure in the HCM it is stated that it does not distinguish between crossing and merging
conflicts. In a crossing conflict vehicles of one traffic stream must traverse the links of another traffic stream (see
Fig. 7 left). A merging conflict describes the situation where two traffic streams use the same destination lane(s).
In the example network a merging conflict exists between the left turns from the major street and the opposing right
turns from the major stream (see Fig. 7 right).
224 Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226
Author name / Transportation Research Procedia 00 (2016) 000–000 9

Fig. 7: Example network with crossing conflict (left) and merging conflict (right)

The HCM procedure determines the conflicting flow rate of 100% regardless of the number of lanes of the exit
link. In VISsim conflicts are modelled in greater detail, i.e. they are considered on a lane by lane basis. That means
actual merging conflicts are only present if the conflicting movements use the same destination lane. Assuming the
southbound exit link in the example network had two lanes and the two movements, left turn from east and right
turns from west, were to use their own lane there would be no conflict between the two traffic streams. In such a
case the HCM would overestimate the conflicting flow potentially resulting in an underestimation of potential
capacity for the left turn movement of rank 2.

Another example is given in Fig. 8 for the minor street right turn movements. In VISsim it is possible to allow
vehicles turning right to use both lanes (see Fig. 8, left), or the offside lane only (see Fig. 8, right).

Fig. 8: Modelling of conflicts for minor RT movement: left turning into both lanes / right turning into the offside lane
only

Right turning vehicle using the offside lane are only in conflict with vehicles on the major street moving in this
lane already. Vehicles moving in the inner lane do not affect vehicles turning into the offside lane. If both lanes
Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226 225

10 Anett Ehlert / Transportation Research Procedia 00 (2016) 000–000


can be used by the right turning vehicles the vehicles in the inner lane are in conflict with those vehicles that turn
into this lane. These right turning vehicles have conflicts with vehicles on both lanes of the major street whereas
vehicles turning into the offside lane are only in conflict with vehicles in the outer lane of the major street.
Consequently, the conflicting flow in the simulation depends on the exact layout of the intersection, i.e. on the
movements by lane.
In contrast, the HCM is using assumptions about the conflicting flow, namely for the right turning vehicles the
conflicting flow consists of 100% of vehicles moving straight on the major street and 50% of vehicles turning right
from the major street in the southbound direction. The latter conflict does not exist in the simulation model and the
conflict with the straight movement depends on the layout and distribution of vehicles across lanes. Consequently,
when calibrating parameters against the HCM assumptions of conflicting flows in the HCM must be taken into
account. Because simulation networks are much more detailed also the conflicting flows can be determined precisely.

5. Conclusion

The calibration exercise for the example network has demonstrated that the parameters provided are suitable to
replicate HCM graphs for situations in which gap acceptance applies and the assumptions used to determine the
conflicting flow match. The default gap values in Vissim provide an excellent starting point for calibration and can
be adjusted where assumptions of the HCM are not fully applicable, either because of local conditions or the
simulation produces more precise data. The parameters provided in Vissim have a direct relationship to the potential
capacity for certain flow levels of the conflicting traffic stream. Fig. 9 shows these effects for the right turn
movement from a minor approach.

Fig. 9: Effect of different junction parameters on the potential capacity

The parameter follow-up gap determines the intersection point with the y-axis, or in other words the capacity at
very low flows on the major traffic stream. The parameter for the critical gap determines capacities at medium flow
levels. The parameter for the maximum waiting time determines how many vehicles from the minor approach can
enter even if the major traffic stream operates near capacity. Moreover, it has been shown that the simulation model
can be more accurate with regard to the actual conflicting flows because intersection layout and the distribution of
traffic flows across lanes is modelled in more detail.
226 Anett Ehlert et al. / Transportation Research Procedia 21 (2017) 216–226
Author name / Transportation Research Procedia 00 (2016) 000–000 11

References
Burghout, W., Koutsopoulos, H., and Andreasson, I. 2006. A discrete-event mesoscopic traffic simulation model for hybrid traffic
simulation, Proc. IEEE Intelligent Transportation Systems Conference, Toronto, Canada
Leonard, D.R., Power, P., and Taylor, N.B.,1989. CONTRAM: structure of the model, Transportation Research Laboratory, Crowthorn
TRL Report RR 178, United Kingdom
Mahut, M., 2001. A discrete flow model for dynamic network loading, Ph.D. thesis, Université de Montréal,
Canada
Nagel, K. and Schreckenberg, M., 1992. A cellular automaton model for freeway traffic, Journal de Physique I France 2 (12), pp. 2221-
2229 PTV Group, 2015, PTV Vissim 8 User manual, Karlsruhe, Germany
Highway Capacity Manual, 2010. Transportation Reseach Board, Washington DC
Wiedemann, R., 1974. Simulation des Verkehrsflusses. Schriftenreihe des Instituts für Verkehrswesen, Heft 8, Universität (TH) Karlsruhe
(seit 2009 KIT – Karlsruher Institut für Technologie)

Вам также может понравиться