Вы находитесь на странице: 1из 6

2014 IEEE International Conference on System Science ang Engineering (ICSSE)

Jnly 11-13 2014,Shanghai, China

An Introduction to Event-based Control


for Networl<ed Control Systems
Qianqian Hong, Hao Zhang*,

*Electronic and Information Engineering Department, TongJi University, Shanghai, China


E-mail: hongqianI102@163.com;zhang_hao@tongji.edu.cn
Abstract-This

paper aims at making an overview of three

event-triggered

control

and

self-triggered

control

different triggering control schemes for networked control

respectively. Many NCS control schemes so far are based

systems based on previous works, namely, time-triggered

on time-triggered communication which makes full use of

control, event-triggered control and self-triggered control .

mature data-sampled algorithm, though there are some

By referring to large amount of others' works, this paper

problems

makes a comparison and analysis of the three triggering

Time-triggered control is an original control method

schemes and introduces some new control methods based on

proposed in last century. It is easy to design and analyze

them. Transmission delay and data dropout are still two

while

serious problems in the networked control systems needing

simultaneously. So event-triggered control and self

due

leads

to

to

limited

network

unnecessary

waste

bandwidth.

of

resources

simply

triggered control methods are developed to solve this

introduced at last. Through the analysis, we can conclude

existing problem afterwards. They successfully reduce the

that event-triggered control method is most people's choice

communication and traffic load in the light of a

and have been used to solve difficult problems in many

nonuniform sampling scheme. These two control schemes

occasions.

make good trade offs between communication load and

to

be

coped

with.

Scheduling

protocols

are

closed-loop
Keywords-event-triggered control; self-triggered control;

which

achieves

the

expected

Moreover, as an extension to event-triggered control

networked control systems


1.

system

performance while occupies less communication load.


scheme, self-triggered control can predict next event time

INTRODUCTION

according to local information and it is not necessary to

Networked control systems have received certain

monitor triggering condition continuously. It needs no

attention in the past decades on account of the fact that

more hardware to monitoring the triggering condition

information

transferring

obviously. Self-triggered control is more flexible than

components

are

between

accomplished

control

through

system
network

communication rather than point-to-point wrings. The


term "Networked Control Systems" first appeared in
Gregory C. Walshs article. In recent years, the analysis
and synthesis of NCSs have been extensively studied in
many works. These networked control systems are in
service throughout the national infrastructure such as
transportation networks and the electrical power grid. A
networked control system always consists of several
components like sensors, controllers, actuators and control
plants which are physically distributed. The usage of
communication network has brought many advantages that
are reduction of costs, higher flexibility and ease of
streamlined installation while maintaining the system
performance. In the triggering control point of view, there
are three main different control methods that are
time-triggered control (periodic data-sampled control),

978-1-4799-4367-8/14/$31.00 2014 IEEE

event-triggered control when triggering an event.


It is known that the networked communication is not as
dependable as a point-to-point communication. Traffic
congestion in a networked control system may lead to
many undesired effects on system stability if severe
enough. In the initial works, many scholars investigated
the control methods for NCSs under a desired network
transmission without consideration of time-delay and
packet dropout such as [8] which made an assumption that
the triggering signal can be transmitted to the input
controller immediately in zero seconds. But in practical
systems, communication delay and data packet dropout are
the most common cases during signal transmission. Based
on the mature analysis of networked control systems,
some control model and control schemes are proposed in
[3], [9] and [18] et al..

The remainder of this paper is organized as follows. Tn

large unnecessary communication in the networked

section IT, a simple overview of time-triggered control is

control systems. Two classes of methods are proposed to

made since it is regarded as a basic control theory for other

cope with the existing problem. The first way is to use rate

control methods. Part ITT will make a review of

adaptation techniques where the sampling period is

event-triggered control and introduction to some novel

determined by the controlled system dynamics or the

methods proposed recently which can solve time-delay

current bandwidth conditions of the system, e.g. [ 14] and

and other serious problems effectively as well as improve

[IS]. The second method is to produce non-periodic

system performance obviously. Tn part TV, self-triggered

executions in the light of event-based data-sampled

control is introduced comparing with event-triggered

techniques introduced in the latter parts. The main goal of

reasonable usage of scheduling

these two methods is to minimize the bandwidth

protocol obeyed by different triggering schemes is also

utilization while still guarantee stability and acceptable

vital in distributed multiple networked control systems.

system performance.

control. Otherwise,

Finally, part V makes a conclusion of triggering control

Ill. EVENT-TRIGGERED CONTROL IN NcSs

schemes in networked control systems.


II.

TIME-TRIGGERED CONTROL IN NcSs

In original control theory researches, open-loop control


scheme is used to realize the control of system under the

Time-triggered control mechanism always leads to


inefficient utilization of networked resources although it
is preferred for us from the analysis and design point of
VIew.

To

mitigate

the

unnecessary

waste

of

optimizing assumption that the control system has

communication and computation resources in traditional

knowledge of initial state and system model and no

time triggered control, event-triggered control scheme has

process disturbance. But there always exists external or

been raised in last century in [ 1] which first certificated

internal disturbance in a system and open-loop control

the potential of this control mechanism to reducing the

scheme is consequently not essential for general system

communication and computation times by experiments

control. Closed-loop control mechanism is more beneficial

and emulation. Soon afterwards the analysis results are

to cope with the uncertainty problem of a system in [32].

derived in [5] which verified that it is able to reduce

Moreover, a better sampling scheme is able to reduce the

information exchange and network bandwidth occupancy

uncertainty property in a more creditable way in NCSs. So

through

event-triggered

control

comparing

with

time-triggered control scheme was proposed in last

time-triggered

century. As a traditional control method it has become the

event-triggered networked control system is showed as

control.

The

basic

structure

of

theoretic basis of event-triggered and self-triggered control

figure 1. Contrast with traditional closed-loop control

schemes, which is also called periodic data-sampled

method, event generator is added between the controlled

control. The sensors sample state data periodically at fixed

plant and controller, which determines whether the

predefined time instances and then transfer sampled

sampled-data should be transmitted or not.

signals to control plants through network communication.

:(f)/)(

A zero-order holder is necessary for time-triggered control


between the sensors and controllers which maintain the

consecutive sampling instants. Many works can be found


last

century

such

as

[33]

printed

-t....iiij ....

1990s.

However, the worst drawback of this approach is that it


makes a fixed usage of the communication bandwidth
regardless of the current network loads or changes in the
control systems. The sampling period is predefined to
maintain the performance stability under the worst
scenario with largest disturbance or traffic congestion.
There always exist a lot of resources wasted on account of

Network

Time-triggered control method is appealing for a long time


on account of its ease to analysis and designing.

Sensor

Event generator

value of input of the plant as a constant between two


m

--,
. I
If}1)(l
I

-------

Plan

Figure 1.

A.

Structure of an event-triggered networked control system

Several Typical Triggering Condition

In initial approaches, some researchers designed the


controllers

under

the

assumption

of

perfect

communication without network transmission delay and


data packet dropout and then determined the maXImum
allowable interval between two consecutive triggering

instances that ensured desired performance of closed-loop


system under a network protocol, such as Round-Robin

e(ikh) An event-based state-feedback approach is proposed


.

(RR) protocol and Try-Once-Discard (TaD). Moreover, in

in [6] which adds a control input generator between the

traditional event-triggered control the system states are

event generator and the control plant. Both generators

monitored continuously to ensure a high degree of

have a same model and update their state with the same

robustness in [ 17]. From the feedback point of view, there

value simultaneously. The former approach is based on an

are two event-triggering schemes respectively based on

assumption that all states are available to us, but when we

state-feedback control and output-feedback control. The

can't get full states information, state-feedback control is

control tasks of these two control mechanisms are both

not useful for us. Then output-feedback control is

executed only if state-dependent triggering condition is

developed by [8] based on Ref. [6]. A Lunberger observer

violated. The simplest triggering condition that often used

is added into the closed-loop system which has knowledge

in many papers can be expressed by

of observed values of system states instead of real states.

Ilx(t)- x(t I

The triggering condition can be designed depending on the


(1)

error between the observed states and event generation

wherex(t)with n dimensions in multi-agent problems or


one dimension for a single plant denotes the real time

states. An algorithm is proposed to design event-based

closed-loop control plant state, while x(t) with same

can refer to Ref. [27] for detail learning about dissipative

dimensions since x (t ) is defined as the state value of the


last triggering time.

is a predefined constant which

indicates the error between x (t) and x(t). We can design a


controller with arbitrary precision by accordingly choosing
the event threshold e . There still exist some other forms of
triggering conditions such as adaptive trigger function
depending on Lyapunov function and time-dependent
triggering function such as taking an exponentially
decreasing variable as threshold [31] et al..

observers for networked control systems in [30]. We also


control based on output-feedback which is designed for
general system.

B.

Monitoring Triggering Condition Discretely

The former control methods all need continuous

supervision of the triggering conditions to ensure the


stability of performance in [6] and [8] and inequality (1) or
(2) are usually used to determine whether event should be
triggered and sampled data should be transmitted or not.
Once the triggering condition deviated from the threshold,
event was executed and the updated signal would be
transmitted

to

the

control

plant

through

network

The right side of inequality (1) is a constant that has no

communication. As figure 1 shows without the sensor in

relationship with the controlled plant, we can substitute

the red dotted line, through which data is sampled

controlled plant for the constant as inequality (2) shows.

periodically. But it is waste of resource to investigate the

Ilx(t) - x(t)1 1

t
b x( )

ll

11

event generator state when the system does't deviate from


(2)

the threshold actually.

When the event generator is investigated as a discrete


manner, we have triggering condition

Periodic event-triggered control (PETC) scheme is


proposed by [29] to solve this problem, where sensors take

(3)
and

sample of the plant state at a constant rate and transfer


sampled information to event generator which determines
whether the event should be triggered. The signal would

(4)
Where

be transmitted to controlled plant only if the triggering


condition is violated from the threshold. Inequality (3) or
(4) are usually used as the triggering condition in discrete
event-triggered control schemes.
C.

tk denotes the last event time, and tk+! denotes next event

time. Inequality (3) and (4) are more reasonable through


the usage of parameters

(5,

<I>

and state-sampled-error

Time-delay And Data Dropouts

All the methods proposed above are available under the


assumption that there are not time-delay and data dropouts.
But in the real networked control systems, these research
results are impractical on account of the existence of

time-delay and packet dropout, which will lead to unstable

method is developed to remit traffic congestion problem in

performance during to the fact that the updated signal and

[24] through adaptive differential modulation.

data packet can't be transferred to system plant in time.

Tn the practical control systems, a networked control

Thus, more control schemes are proposed afterwards

system is usually consisted of several multiple control

based on previous studies taking into account transmission

loops which contain plants, sensors and actuators,

delay and packet dropout, for example, in [26].

respectively. It is intuitively called distributed networked


studied

control systems (DNCSs). Centralized event-triggered

communication delay and packet dropout separately while

control is impractical for large scale networked control

Tn

earlier

works,

some

researchers

some others considered the two issues together. Based on

systems.

PETC scheme, Ref. [3] proposed a new algorithm to solve

scheme where a subsystem is able to determine event

time-delay and data-packet loss problems that develop a

instants using its own locally sampled data in [ 13]. One

fundamental

distributed

event-triggering

new event interval model by using virtual interval

can refer to more works about this direction such as [4].

sampling technique that the ZOH interval is divided into

Each agent determines its own event according to its own


state and its neighboring

several subsets according to sampling period.

agents'

information

that

connected with it. All these works are encountered a


If II
'en or

node

If there is no packet loss

- - - - - 11(ExaIlJP' (= 4f

11

- - - -,J

common difficult problem that the interaction between

III

agents is limited. Event-based model predictive control is


proposed to solve this problem successively in [25].

Conlroller node i
ZOH ubSel'!lr

! 1
1"1------

. -

_ _ _,
,
,

'111 + Tt,
Figure 2,

112
-- -

I
?I

n'l
---

O H old!liJc,".!
_

{l,

-----

I.

IV. SELF-TRIGGERED CONTROL IN NcSs

-""9

Self-triggered control model was firstly coined by [11]

III t

rto..

in the context of real-time systems for linear system.


Aperiodic sampling control scheme in the networked

subsets of ZOH holding interval

Contrasting with the algorithm proposed in [8] which


determined triggering condition and state-feedback gain in
two steps, the new algorithm was developed to design both
of them in one step for convenience while maintained

control system is proposed by [ 12] based on theory in the


Ref. [11].
A,

Comparing With Event-triggered Control

As an alternative to event-triggered control, self

performance of closed-loop system. It shows that we can

triggered control has some considerable advantages in

determine

system

contrast with the former. It is not necessary to monitor the

performance we want to obtain. This method is more

triggering condition continuously so that this triggering

flexible. A similar control method designing triggering

mechanism can provide additional resource reduction for

condition and output feedback controller together was

the sensors and also less complexity in the physical

developed in [28], where a delay system model that took

implementation in [19]. Moreover, it provides a more

controller

gain

according

to

account into time delays and data packet dropouts together

flexible way to determine trigger instant which possibly

was firstly proposed for the purpose of system analysis

leads to no executions if the system is stable at its

and control design. Ref. [ 16] proposed a new algorithm

equilibrium point. Moreover, it needs no more specific

using the definite difference approximation for the delay

extra hardware to complete unremittingly supervision of

term to estimate the maximum allowable delay in a NCS.

triggering condition in contrast with event-triggered

The control input generator and event generator are

control. The attractive advantages of self-triggered control

designed by a same model in Ref. [2]. Thus the control

lead to a prevalent trend to investigate self-triggered

input generator states would be updated according to states

control

of event generator when time-delay and data dropout

corroborates the strong robustness of self-triggered control

happened. This method is easy to understand and apply for

with m ! * Ui71 ffl ${

us in physical design. Otherwise, traffic reduction is a

mechanism.

The

previous

work

[22]

stability to transmission

delay in networked control system in the usage of system

permanent topic for us to consider, all the former works

simulation. An implementation of self-triggered controller

have been trying to improve this property of a NCS.

in networked control systems is described to certificate

Recently, event-triggered adaptive differential modulation

that self-triggered control scheme can really be used for

reducing bandwidth occupancy while maintain stable

triggering schemes property, achievable performance

performance in [ 13] through an experimental case.

under global and local scheduling decisions et al.. One can

B. New Control Schemes Based On Self-triggered


Control

refer to [10] and Ref. [7] gives a specific introduction to

There exist several common points when we make an

analysis

of

stability

and

functions

as

control and event-triggered control regardless of specific


control schemes. Time-delay model is established in [34]
and an optimizing controller is designed to complete the
implementation. But this triggering condition is designed
for some special systems. Based on the former works on
event-triggered control, self-triggered model predictive
control with delay compensation is proposed in [20] by the
usage of two kinds of control inputs, namely the standard
control input and the delay compensation input which can
be obtained by solving the optimal control problem. A
special self-triggered control scheme is developed in [2 1]
for two classes of nonlinear system in networked control
system, namely, polynomial systems and state-dependent
homogeneous systems. The trigg- ering condition that
depends on the current measurement of states, the desired
performance and the dynamics of the system is simpler
than that are proposed in previous works. This triggering
condition has broad usage and can easily be applied to
similar

systems.

basic

introduction

to

self-triggered control in distributed networked control


systems is made in [23] and one can refer to it for detail
knowledge.

C.

In network communication, a new data packet packet

(DPP) containing all physically distributed sensors and


actuators' data packets is used to generate a new signal
through

network.

The

so-called

DPP

transmission problem has attracted certain attention so far,


which generally implies the collocation of all sensors and
actuators perfectly. Nevertheless, such an assumption does
not hold in many cases. It is essential to determine a
scheduling protocol obeyed by the central coordinator to
select the suitable data transmission node whenever a
predefined triggering rule is satisfied. Thus it is important
to make an analysis of scheduling protocol for with
different triggering schemes in networked control systems.
In

general,

V.

CONCLUSION

This paper mainly makes an overview of three different


triggering control methods from the points of triggering
schemes,

occupancy

of

system

bandwidth

and

performance stability and robustness et al. for NCSs.


Many works have been done so far, and the triggering
control

methods

are

considerable

mature.

While

time-triggered control is regarded as a basic knowledge


and essential component in modern NCSs, most are eager
to investigate event-triggered control on account of its
good properties and broad usage. Self-triggered control
has better control properties theoretically, but in fact it is
not easy to monitor each agent and keep the consensus of
them. This kind of control needs large amount of signal
communication among distributed agents.
Though many controllers developed in previous works
can be implemented in a relatively stable way, most of the
schemes are realized under some certain conservative
assumptions such as maximum allowable packet dropout
bound condition and without packet disorders. We can
also make extension researches in the following two
aspects:
(1) Most stability analysis algorithms are proposed as a
sufficient condition for the controller or triggering

Scheduling Protocols

transmitting

scheduling scheme for event-triggered control.

network

communication load reduction between self-triggered

other

Specific analysis is made through comparing different

there are several scheduling

protocols

including TDMA (time division multiple access), FDMA


(frequency division multiple access), and CSMA (carrier
sense multiple access) to multiplex the data streams in [ 10].

condition, even some of which are strongly conservative


that can be only applied to one fixed models or systems.
These triggering schemes are only limited to theoretical
analysis that has no broad applications. What we can do in
the future is trying to release the essential conditions such
as tolerating longer transmission delay time or larger
packet dropout amount and find some more available
algorithms or control models.
(2) We can seek for other new control methods that can
be combined with triggering control. Recently some
researchers begin to applying fuzzy logic control scheme
to NCSs. This is a relatively new method that is immature
for networked control systems. It is easier for us to find
some new methods in the light of fuzzy logic control
consequently.
VI. ACKNOWLEDGEMENT
This work is supported by the National Natural Science

Foundation of China (6 1273026) and the Fundamental


Research Funds for the Central Universities.

K.

E.

Arzen, "A

simple

event

PID

controller,"

in

and

J.

Lunze,

"Event-based

control

with

intlnity control co-design for networked control systems," vol.

event-triggered

control

with

network

delays and packet losses," 51st IEEE Conference on Decision and


Control,pp. 1-6,2012.
event-driven controllers for linear systems," International Journal
of Control,vol. 81(4),pp. 571-590,2007.

[7]

Lunze

and

D.

Lehmann, "A state-feedback approach

to

Decision and Control,pp. 6716-6721,2010.


[24] P. Upeka, K. H. Saman and M. Y. M. Iven, "Event triggered
differential

modulation:

new

method

for

traffic

reduction in networked control systems," IEEE Transactions on


Automatic Control,vol. 58(7),pp. 1696-1706,2013.
predictive control for the cooperation of distributed agents,"
American Control Conference,pp. 6473-6478,2012.
[26] S. M. Magdi and I. Abdulla, "Role of delays in networked control
the 2003 10th

IEEE International

Conference on Electronics, Circuits, and Systems, vol. I, pp.


40-43,2003.
[27] J. Wang, Q. L. Han and F. W. Yang, "Event-triggered output
feedback dissipative control for network-based systems," 39th

Annual Conference of the IEEE Electronics Society, pp. 5086-

P. tabuada, "Event-triggered real-time scheduling of stabilizing

5091,2013.

D. Lehmann and J. Lunze,"Event-based output-feedback control,"


19th Mediterranean Conference on Control and Automation, pp.
982-987,2011.

[9]

sample:

event-based control," Automatica,vol. 46(1), pp. 211-215,2010.


control tasks,"IEEE Transactions on Automatic Control, vol. 52
(9),pp. 1680-1685,2007.
[8]

to

Control,vol. 1(1),pp. 15238-15243,2008.

systems," Proceedings of

W. P. M. H. Heemels,J. Sandee and P. V. D. Bosch, "Analysis of

J.

not

[25] E. Alina, V. D. Dimos and J. K. Kostas, "Event-based model

M. Guinaldo, D. Lehmann, J. Sanchez, S. Dormido and K. H.


Johansson, "Distributed

[6]

or

[23] V. D. Dimos, F. Emilio and H. J. Karl, "Distributed self-triggered

adaptive

P. Chen and C. Y. Tai, "Event-triggered communication and H


49(5),pp. 1326-1332,2013.

[5]

sample

control for multi-agent systems," 49th IEEE Conference on

Lehmann,

Control,vol. 85(5),pp. 563-577,2012.

[4]

Tabuada, "To

17th World Congress, International Federation of Automatic

based

communication delays and packet losses, "International Journal of


[3]

P.

feedback control systems with L2 stability," Proceeding of the

1999.
D.

and

Automatic Control,vol. 55(9),pp. 2030-2042, 2010.

Proceeding14th IFAC World Congress, vol. 18, pp. 423-428,


[2]

Anta

selt:triggered control for nonlinear systems," IEEE Transactions


[22] X. F. Wang and M. D. Lemmon, "State based selt:triggered

REFERENCE
[I]

[21] A.

[28] X. M. Zhang and Q. L. Han, "Event-triggered mixed H infinity


and

passive

feedback,"

control

39th

of

linear

Annual

systems

Conference

of

via

dynamic

the

IEEE

output

Industrial

Electronics Society,pp. 5080-5085,2013.


[29] W. P. M. H. Heemels, M. C. F. Donkers and R. T. Andrew,

X. F. Wang and M. D. Lemmon, "Event-triggering in distributed


networked control systems," IEEE Transactions on Automatic
Control. Vol. (3),pp. 586-601,2007.

"Periodic

event-triggered

control

for

linear

systems,"

IEEE

Transactions on Automatic Control,vol. 58(4),pp. 847-861,2013.


[30] M. Pablo, T. Ubaldo, F. Carlo, V. D. Dimos, H. J. Karl and R.R.

[10] C. Anton and H. Toivo, "Scheduling of event-triggered controllers


on a shared network," Proceeding of the 47th IEEE Conference on
Decision and Control,pp.3601-3606,2008.

Francisco, "Distributed event-based observers for LTI networked


systems," 2012.
[31] D.

[II] M. Velasco, P. Mart and J. M. Fuertes, "The self-triggered task


model for real-time control systems," in WIP of the 24th IEEE
Real-Time Systems Symposium,2003.

V.

Dimarogonas, K.

H.

Johansson

and

F.

Allgower,

"Event-based Control for Multi-Agent Systems," Master's Degree


Project,Stockholm,2010.
[32] K. Y. You and L. H. Xie, "Survey of Recent Progress in

[12] A. Anta and P. Tabuada, "On the benefits of relaxing the


periodicity assumption for networked control systems over CAN,"
30th IEEE Real Time Systems Symposium,pp. 3-12,2009.

Networked Control Systems," Acta Automatica Sinca, vol. 39(2),


pp. 101-118,2013.
[33] H. M. AI-Rahmani and G. F. Franklin, "A new optimal multi-rate

[13] C. Antonio, M. Pau, V. Manel, L. Camilo, V. Ricard, M. F. Josep

control of linear periodic and time-invariant systems," IEEE

and G. Eullia, "Self-triggered networked control systems: an

Transactions Automatic Control,vol. 35,pp. 406-415,1990.

experimental case study," Industrial Technology (lClT), 2010

[34] F. L. Lian, J. R. Moyne,and D. M. Tilbury, 'Time-delay modeling

IEEE International conference,pp. 123-128,2010.

and optimal controller design for networked control systems,"

[14] H. Rehbinder and M. Sanfridson, "Scheduling of a limited


communication channel for optimal control," Automatica, vol.
40(3),pp. 491-500,2004.
[15] M. E. M. Ben Gaid, A. Cela and Y. Hamam, "Optimal integrated
control

and

scheduling

communication

of

constraints:

networked

control

application

to

systems
car

with

suspension

system," IEEE Transactions on Control Systems Technology, vol.


14(4),pp. 776-787,2006.
[16] A. F. Khalil and J. H. Wang, "A new method for estimating the
maximum allowable delay in networked control of bounded
nonlinear systems,"17th International Conference on Automation
and Computing,pp. 80-85,20II.
[17] P. Romain, P. Tabuada, N. Dragan and A. Anta, "Event-triggered
and self-triggered stabilization of distributed networked control
systems," 50th IEEE Conference on Decision and Control and
European Control Conference ,pp. 2565-2570,2011.
[18] D. Yue, E. G. Tian, and Q. L. Han, "A delay system method for
designing

event-triggered

controllers

of

networked

control

systems," IEEE Transactions on Automatic Control,vol. 58(2),pp.


475-481,2013.
[19] X. Wang and M. Lemmon, "Self-triggered feedback control
systems

with

finite-gain

L2

stability,"

IEEE

Transactions

Automatic Control,vol. 45(3),pp. 452-467,2009.


[20] K. Koichi and H. Kunihiko, "Self-triggered model predictive
control with delay compensation for networked control systems,"
38th Annual Conference on IEEE Industrial Electronics Society,
pp.

3200-3205,2012.

International Journal of Control,vol. 76(6),pp. 591-606,2003.

Вам также может понравиться