A Layered Multi-Agent Model for Multi-Conﬁguration Platoon Control

Baudouin Dafﬂon, Franck Gechter, Pablo Gruer and Abderraﬁaa Koukam

IRTES-SET, UTBM, Belfort, France

Keywords:

Multi-Agent Model, Platoon Control.

Abstract:

Nowadays, urban environments suffer from recurrent trafﬁc jam with associated side effects. Platoon system,

a set of vehicle attach by virtual link, is one of answer. in order to cope with issues such as different platoon

geometrics since several years, we’re seeing the multi-agent system as element of response to many problems.

This paper to apply this concept for platoon vehicles system through an multi-agent model based on key layers.

Thanks this solution, implemented as an agent which makes decisions depending only on its own perception

where each vehicles is recorded, analysed, transform in car command and adapted to vehicle by a command

ﬁlter.

1 INTRODUCTION

Platoons can be deﬁned as a set of autonomous vehi-

cles evolving on a particular environment while main-

taining a particular geometric conﬁguration. Platoon

control consists in determining the behaviour of each

one of the vehicles during platoon evolution, in order

to maintain the conﬁguration and adapt it to changes

in the terrain (presence of unanticipated obstacles, de-

creasing of available surface, ...).

Several international projects, past or present, address

platoon control. Among them, we can cite PATH

(Hedrick et al., 1994), SARTRE (Chan et al., 2011),

CRISTAL

, SAFEPLATOON (Cartade et al., 2002)

. Most of them deal with column platoons as unique

conﬁguration and situate in well deﬁned environment

such as highways where the curve radius are high and

where the speed can be considered to be constant most

of the time. However, other application domains,

placed in different kinds of environments could bene-

ﬁt from platoons composed of different types of vehi-

cles. Among those application domains we can men-

sion transportation and maintenance operations in ur-

ban areas, labouring and harvesting in agricultural ar-

eas and military operation theatres. Those cases are

subject to diverse, more stringent constraints. As an

example, in urban areas with a column conﬁguration,

the lateral error must be highly limited in order to

avoid collisions. In echelon or line conﬁgurations, en-

vironment related constraints are more inﬂuential.

http://projet-cristal.net/

http://web.utbm.fr/safeplatoon

In this paper, we propose an multi-agent based ap-

proach for the multi-conﬁguration platoon control

problem. The approach can be considered as self-

organizing, because platoon conﬁguration emerges

from the behaviour of each vehicle, strictly based on

local vehicle’s perceptions. The platoon’s conﬁgura-

tion is determined locally by assigning to any platoon

vehicle another neighbouring platoon vehicle, consid-

ered as local leader. This proposal is structured as

a multi-layer decision process dealing ﬁrst with the

interpretation of the perception data, then integrating

the choice of the local leader depending on the in-

tended spatial conﬁguration and ﬁnally producing a

kinematic decision that has to be performed by the

vehicle. In this approach, each vehicle is consid-

ered as an agent which bases on its perceptions and

on the local intended conﬁguration to make the cor-

rect decision. The proposed approach integrates ob-

stacle avoidance abilities which is based on a multi-

agent ﬁltering method similar as the one developed in

(Franck Gechter and Koukam, 2010).

The paper is structured as follow: after a short re-

minder about vocabulary, a state of the art on platoon

systems is proposed. Then, the multi agent system ap-

plied to platoon control is described. Finally, a con-

clusion and some considerations on future work di-

rections are presented.

Dafﬂon B., Gechter F., Gruer P. and Koukam A..

A Layered Multi-Agent Model for Multi-Conﬁguration Platoon Control.

DOI: 10.5220/0004632500330040

In Proceedings of the 10th International Conference on Informatics in Control, Automation and Robotics (ICINCO-2013), pages 33-40

ISBN: 978-989-8565-70-9

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

2 DEFINITIONS

The literature introduces a rich terminology in related

to the platoon domain. In this section, we intend to

present the vocabulary used along the work, in order

to avoid ambiguities.

2.1 Leaders

Vehicles with a distinctive role within the platoon are

frequently qualiﬁed as leaders. We distinguish two

kind of leader roles.

Global Leader. The global leader is the reference ve-

hicle of the entire platoon. It can be fully autonomous,

applying a path-following algorithm, or driven by a

human operator. The global leader determines the ref-

erence trajectory for the convoy. Local Leader. The

local leader notion is tied to local, self-organizing ap-

proaches, and corresponds to the vehicle taken as a

reference by a follower vehicle. Each vehicle in the

platoon has a local leader, but this role can be assigned

dynamically during platoon operation. Generally, the

local leader is taken among the closest vehicles in the

follower perception ﬁeld.

Virtual Leader. The notion of virtual leader is tied to

the mechanisms involved in our methods. The princi-

ple developed in this paper, is to be able to transform

any spatial conﬁguration into a local column conﬁg-

uration of a local leader and its follower, and to ap-

ply a well deﬁned interaction model, to determine fol-

lower’s behaviour.

2.2 Geometry

A platoon conﬁguration geometry is deﬁned by means

of two distances, lateral and the longitudinal (cf.

Fig.1). The Lateral distance represents the lateral

spacing between two neighbour vehicles. The Lon-

gitudinal Distance. represents the spacing between

two neighbour vehicles, in the direction of the move.

Figure 1: Lateral and longitudinal distances.

Conﬁguring a platoon formation bases on the deﬁ-

nition of both lateral and longitudinal distance. De-

pending on the values of these, several platoon con-

ﬁguration can then be deﬁned, among which we can

mention: Column Conﬁguration. This conﬁgura-

tion, represents the most frequently studied form of

platoon where vehicles organize as train (cf.ﬁgure

2). In this conﬁguration, lateral distance should re-

main as small as possible (for curved trajectories this

means mono-trace displacement). Column conﬁgu-

rations have been foreseen as mostly dedicated for

the transport of passengers in urban or highway trans-

portation systems.

Line Conﬁguration. In this conﬁguration, vehicles

are placed side by side (cf.ﬁgure 2). The longitudinal

distance must be null. This conﬁguration can be ap-

plied to agricultural tasks, such as tilling.

Echelon Conﬁguration. Vehicles are in an inclined

column conﬁguration, each is offset from the preced-

ing by a lateral distance (cf.ﬁgure 2). In this conﬁgu-

ration, lateral and longitudinal nominal distances have

a speciﬁed, non null value. This conﬁguration can be

dedicated to agricultural or military applications.

Arbitrary Conﬁguration and Wedge Conﬁgura-

tion. Arbitrary conﬁgurations can hold many geomet-

rical forms, produced by the combinations of two or

three of the conﬁgurations described above. These

conﬁgurations are mostly used in military environ-

ments. In wedge conﬁguration for instance, leader

vehicle is followed by echelons of vehicles placed to

the right and the left forming an inverted ”V” forma-

tion (cf. ﬁgure 2).

Figure 2: Conﬁgurations of platoon. From left to right: col-

umn, line, echelon, wedge.

3 STATE OF THE ART

Platoon control approaches can be sorted into two

main categories: local approaches base on a local ref-

erence frame. On the other hand, global approaches

base on a global frame, common to every vehicle in

the platoon.

3.1 Local Approaches

Local approaches are based on a local reference frame

which generally anchors in the follower vehicle. In

this frame, the position and the orientation of the local

leader can be determined in the local frame. Conse-

quently, the local approach can be considered as the

ICINCO2013-10thInternationalConferenceonInformaticsinControl,AutomationandRobotics

regulation of both lateral and longitudinal distances

between the follower and its local leader.

Among the proposals corresponding to this ap-

proach, we can mention:

• System based on an automatic control mecha-

nism: (Daviet and Parent, 1996) proposes a mech-

anism for local control based on a PID (Propor-

tional, Integral, Derivative) controller . The mea-

surement acquisition is performed by linear cam-

eras or by a range ﬁnder sensor (Riess, 2000). The

control reference is decomposed on one hand in a

longitudinal reference and on the other hand in a

lateral reference.

• Control based on undumped impedances: (Gehrig

and Stein, 2001) presents a model of light links

where vehicles are treated as particles subjected

to physical forces.

• Control model based on a double impedance con-

trol: in (o Yeong Yi and Chong., 2005) an

impedance control model is used as a model of

the immaterial link between vehicles.

• Control based on a virtual mechanical link: In

these approaches, each vehicle in the convoy is

designed as an intelligent system able to perceive

its environment and maintaining a pre-set distance

with the preceding vehicle. The platoon system is

the result of direct interaction between each vehi-

cle and its predecessor. This approach is based on

an interaction model inspired by the physics. In-

deed, the interaction between two successive ve-

hicles virtual link shown by a mass-spring type.

Among these one can cite (Franck Gechter and

Koukam, 2010).

3.2 Global Approaches

Each vehicle determines its control references de-

pending on a global shared reference. Platoon trajec-

tory is determined by the global leader and expressed

as a series of trajectory points situated in the global

reference frame, known by every vehicle. The shar-

ing of the trajectory points implies vehicle to vehi-

cle communication capabilities and high performance

global localisation devices aimed at determining the

position of each vehicle in the global reference frame.

Platoon control can be considered as the regulation of

both lateral and longitudinal distance between each

vehicle and the reference trajectory.

Among the proposals corresponding to this ap-

proach, we can mention:

• An optimal controller is proposed in (Levine and

Athans., 1966) to serve on a constant set of inter-

vehicle distance in columns moving at high speed.

• In (Caicedo et al., 2003), the formation is char-

acterized by a set of generalized coordinates de-

scribing the position and the orientation. The re-

sulting shape and evolution of the column is based

on the laws of mechanics.

Relatively to military platoons enumerated, these

can ﬁt the global/local classiﬁcation. Unit center ref-

erenced and leader referenced approaches can be con-

sidered as global (the control can be centralized or de-

centralized). By contrast, Neighbour-referenced tech-

nique is local and decentralized.

4 MULTI LAYERS DECISION

PROCESS

4.1 Global Overview

Our proposal is based on a multi-layer systems for

decision making. This architecture provides a set of

5 plug and play units (cf. Fig 3). The principle de-

veloped is based on the transformation of any spatial

conﬁguration into a column conﬁguration which uses

a well deﬁned interaction model. Then, we use a vir-

tual leader, the position and the orientation of which

correspond to the transformation of the local leader

position and orientation taking into account platoon

spatial desired conﬁguration.

Figure 3: Multi-layer Architecture.

Each step of the process is independent and has got

speciﬁc parameters and can be deﬁned as follow :

Perception. is an abstraction of sensors. The in-

put data are the raw information from the sensors. A

parser is used to convert them in workable data set for

further decision-making process

In Perception Filter. block a ﬁltering policy is ap-

plied. The goal of this unit is to sort out the perceived

elements between obstacles and other convoy vehi-

cles. Among these vehicles a local leader is chosen

using speciﬁc strategy.

The Multi-Conﬁguration. unit applies the geomet-

rical transformations required to change local leader

position and orientation into virtual leader ones.

ALayeredMulti-AgentModelforMulti-ConfigurationPlatoonControl

The Decision-Making. block corresponds to the ap-

plication of the interaction model and the computing

of the command law to be applied to the vehicle.

The last step in this process is an Command Filter-

ing. which can corresponds to driving assistance ﬁl-

tering such as obstacle avoidance or to the introduc-

tion of a kinematic model for the smooth command

for instance.

4.2 Detailed Description

This section aims at describing layers one by one fol-

lowing the introduction description and ﬁgure 3.

4.2.1 Perception

The perception takes as an input the raw data of the

sensor associated to object detection algorithm. To be

able to have reliable detection, sensors have to cover

the surrounded environment of the vehicle as shown

in ﬁg. 4. The output of the perception unit is a list of

localized object in the vehicle reference frame. Vari-

ous kind of sensors can be used to performed this pro-

cess such as stereo cameras, sonars belt, laser range

ﬁnders,...

Figure 4: Sensors ideal coverage area.

4.2.2 Perception Filter

The goal of this unit is to analyse the output of the

perception in order to remove noise in the data and to

sort out detected objects. A Kalman ﬁlter has been

chosen to ﬁlter the noise. A Kalman ﬁlter is a pre-

diction/correction based ﬁlter which uses a transition

model for the prediction and an observation model for

the correction. As for the sorting of the objects a sim-

ple strategy has been used. the aim is to provide a

classiﬁcation of the environment in 3 class.

1. Object member of convoy

2. Object moving in same direction but not member

of convoy

3. Object moving perilously.

This classiﬁcation is based on dynamic study of per-

ceived object.

Then, the ﬁnal step of this ﬁltering unit is a geomet-

rical transformation aimed at expressing the coordi-

nates of the detected objects into the vehicle reference

frame. (Initially, their coordinates are expressed into

the sensors reference frames)...

4.2.3 Multi-Conﬁguration

As exposed before, each vehicle in the platoon can

be seen as an agent that acts based only on its per-

ceptions. For each agent, we deﬁne a leader among

its neighbours in the platoon (see section 4.2.2). The

agent computes its references based on the position of

its leader by trying to maintain the desired lateral and

longitudinal spacing and the correct orientation. Col-

umn formation platoon control functions are now well

known and expose reliable properties. Consequently,

it as been decided to base our approach on this el-

ementary function. So, the key step is to translate

leader local position in vehicle agent reference frame

in order to be able to use column platoon function and

integrates desired lateral and longitudinal distances.

As exposed in section 2, there are several types of

conﬁguration that can be grouped into several fam-

ilies (echelon, line, column). The multi conﬁgura-

tion proposal is based on modiﬁcation of perception

by introduction of virtual leader vehicle. Indeed, as

detailed in table 5, every formations are possible with

the introduction of a virtual leader.

Figure 5: Formations tab.

In order to determine position of virtual leader, sev-

eral parameters are used: geometrics aspects, percep-

tion and conﬁguration order. Indeed, a translation of

la and lo is made in the referential of local leader.

Position of virtual leader is deﬁned by a 2D Trans-

formation (ﬁgure 6) could be describe by following

equation :



= x

+ l

cos(θ) +l

sin(θ)

= y

− l

sin(θ) + l

cos(θ)

(1)

where

ICINCO2013-10thInternationalConferenceonInformaticsinControl,AutomationandRobotics

• (x

) virtual leader position

• (x

) real leader position

• (l

) lateral and longitudinal distances

• θ angle between cars.

Figure 6: Virtual leader in multi conﬁguration.

4.2.4 Interaction Model

A model of interaction inspired by physics has been

chosen. This model, detailed in (El-Zaher et al.,

2011b) has got two springs and a damper placed

between the local leader and the follower vehicle

(cf Figure 7). This model is virtual and unidirec-

tional. Indeed to better control the interaction, the lo-

cal leader vehicle is not affected by generated forces.

Figure 7: Interaction model.

The parameter of this model take into account the

vehicles characteristics . This optimisation allow to

keep a low latency and keep a real-time reactivity.

A compositional veriﬁcation of this system has been

made in (El-Zaher et al., 2011a) to prove non collision

event.

4.2.5 Command Filter

The command proposed by the interaction model

must be adapted to the environment in order to deal

with the presence of obstacles for instance. Several

ﬁlters can be considered. The most simple is an emer-

gency stop ﬁlter: in case of obstacles in a neighbour-

hood too close, the command is set to 0. Another

ﬁlter presented in previous work (Franck Gechter

and Koukam, 2010) and aimed at computing obsta-

cle avoidance strategies can be apply. This obstacle

avoidance device is based on a multi-agent system

where the observation of an agents population lead

to a modiﬁcation of the vehicle command.

These ﬁlters require data from the perception.

This is why a link is possible to bring directly sen-

sors data from perception unit (or perception ﬁlter) to

command ﬁlter as shown in the picture 8.

Figure 8: Command ﬁlter input.

4.3 Flexible and Adaptive System

As seen above, our proposal is based on layers. Each

module takes, as an input, the output of preceding

block. Thus, provided that the input and the output are

correct, layers can be changed using other algorithms

or strategies. Moreover, it is possible to combined

some inputs or to combine in cascade several blocks

of the same type. Finally, these changes in blocks can

be performed in runtime

Here are two examples to illustrate this modular-

ity:

1. Input combined and association in cascade

Figure 9: Input combined.

2. Interchanged blocks:

Figure 10: Association in cascade.

ALayeredMulti-AgentModelforMulti-ConfigurationPlatoonControl

5 EXPERIMENTAL RESULTS

5.1 Global Overview

5.1.1 Vehicles Description

Vehicles used in the simulation represent (graphi-

cally and physically) the experimental laboratory’s

vehicles. They satisfy the physical constraints and

share the same characteristics. In simulation, they

are equipped with two 270 degrees virtual laser range

ﬁnder, replica of LMS SICK 200 and with GPS-

RTK simulation required to follow and study trajec-

tories. Vehicles have the following characteristics:

1.8m width, 3.05m length, max steering angle=30 de-

grees and max speed=12m/s

5.1.2 Simulator

To assess the quality of our approach, simulations

have been done using VIVUS simulator (Lamotte

et al., 2010), VIVUS is a vehicle simulator based on

PhysX for real physical behaviour developed by the

SeT

laboratory.

This software can simulates behaviours for each ve-

hicle such as perception with laser range ﬁnder or

cameras, physical reaction between elements (wheels,

car’s parts,...),... Physical reaction are computed us-

ing the same physical law as real world (collision,

gravity,...) and considering the peculiarity of the en-

vironment (friction with soil, materials of soils and

walls,...). VIVUS has already been used to test vari-

ous intelligent vehicle algorithms such as linear pla-

toon control (Franck Gechter and Koukam, 2010) and

(El-Zaher et al., 2011b), obstacle avoidance and driv-

ing assistance (Gechter et al., ) and (Dafﬂon et al.,

2012), and intelligent crossroads simulations in (Daf-

fon et al., 2011).

5.1.3 Metrics and Analysis

Two informations are measured during experimenta-

tions:

• Lateral distance measures the spacing between the

horizontal axes of two neighbour vehicles. In

cases of column platoon, this distance should be

null.

• Longitudinal distance represents the inter-vehicle

distance between two neighbour vehicles.

Lateral and longitudinal distances are recorded by

VIVUS simulator and analysed off-line by matlab

scripts. thanks to these measures, we will explore:

http://set.utbm.fr/

- General behaviour of the convoy route with GPS -

Lateral and longitudinal distances a function of time -

Lateral and longitudinal distances a function of speed.

5.1.4 Test Area

Simulations were performed on a 3D geo-localized

model of the city of Belfort (France). Two different

trajectories have been chosen. In the ﬁrst trajectory,

vehicles have to follow a circle around 25m of radius.

The simulations are done many times (ie: 500 itera-

tions) for the purpose a statistical studies about exper-

imentations.

5.2 Experimental Results

Experimentations are made following two scenarios.

The ﬁrst simulates a convoy of vehicles in the column

whereas the second simulates a echelon convoy.

5.2.1 5 Vehicle Platoon Convoy in Column

1. 5 vehicle platoon convoy in line :

In Figure 11, only the lateral gaps from the ﬁrst,

third, fourth and last follower are presented. We

can see that the error between the measurement

and instruction is ampliﬁed depending on the po-

sition of the vehicle in the convoy. The last fol-

lower error is between 0.07m and 0.55m, while

the ﬁrst follower error is between 0.12m and

0.01m. The difference between the measurement

and setpoint comes from ampliﬁcation errors due

to the local approach. It is interesting to compare

these measures to the width of a tire (about 0.2 m).

Figure 11: Lateral Gap in Platoon Convoy.

In Figure 12, only the longitudinal gaps from the

ﬁrst, third and last follower are represented. We

can see that the error between the measurement

and set point is ampliﬁed. The error of ﬁrst fol-

lower is between 5.7m and 6.7m whereas the last

follower is between 5.2 m and 6.4 m. The differ-

ence between the measurement and the reference

is from an extension of the springs in the present

interaction model.

2. 5 vehicle platoon convoy in echelon

Figure 13 shows the lateral deviation follower 1

and 5. We notice that the echelon formation does

ICINCO2013-10thInternationalConferenceonInformaticsinControl,AutomationandRobotics

Figure 12: Longitudinal Gap in Column Platoon Con-

voy.

not lead additional oscillation as compared to a

column formation. The virtual vehicle seems a reliable

alternative for the inter-distance management of a

multi-lateral conﬁguration.

Figure 13: Lateral Gap in Echelon Platoon Convoy.

Figure 14 shows the longitudinal inter-distance mea-

sured on the train. We can notice that the error between

the measurement and set point spreads by amplifying.

The inter-distance of ﬁrst follower is between 5.18m

and 5.2m while the last follower is between 5.01m and

5.22 m.

Figure 14: Longitudinal Gap in Echelon Platoon Con-

voy.

5.2.2 5 Vehicle Platoon Convoy in Circle

1. 5 vehicle platoon convoy in column

As shown in Figure 15 the differences between

ﬁrst and last follower, we can see besides the

residual oscillations that the differences between

vehicles are of the same order as in previous ex-

periment. From the ﬁrst follower to the last,

the longitudinal gap is in a range from 5.87m to

6.04m.

Figure 15: Lateral Gap in Column Platoon Convoy.

Figure 16 represents the lateral error from the ﬁrst

follower to the last. Thus, while the measurement

of the ﬁrst Follower are in the same order as be-

fore, the last follower describes an lateral error be-

tween -0.1 and 0.2m.

Figure 16: Longitudinal Gap in Column Platoon Con-

voy.

2. 5 vehicle platoon convoy in echelon

Figure 17 describes the evolution in time of the lateral

deviations. it may be noted that the position in the con-

voy does not increase the oscillations. Indeed, with a

circular path, the error is comprised between 0.09 and

0.11m (less than the width of a tire). It is interesting to

note that the train keeps the same stability with 3 to 5

vehicles.

Figure 17: Lateral Gap in Echelon Platoon Convoy.

Figure 18 shows that the most the vehicle is near the

center, the more the error on the longitudinal set point

is large. The reasons of this result are the links between

anticipation, speed elongation and virtual vehicle trans-

formation.

ALayeredMulti-AgentModelforMulti-ConfigurationPlatoonControl

Figure 18: Longitudinal Gap in Echelon Platoon

Convoy.

6 CONCLUSIONS

The paper presents a agent approach for platoon sys-

tem through a generic and modular decision process

for autonomous vehicle in platoon system. In this

model, different layer is are proposed and combined

to deﬁne behaviour. Each layer allow facilitate the

processing done in next step. The main advantages

proposed by this system is a run time and self adap-

tation to environment. This solution was successfully

tested in simulation and results obtained are encour-

aging to test using real laboratory vehicles and real

sensors. In order to continue this research, we are now

working on generic and emerging perception ﬁlter to

adapt perception to any sensors.

Those works are done with the support of the

French ANR (National Research Agency) through the

ANR-VTT SafePlatoon

project (ANR-10-VPTT-

011).

REFERENCES

Caicedo, R., Valasek, J., and Junkins, J. (2003). Preliminary

results of one-dimensional vehicle formation control

using structural analogy. Control Conference.

Cartade, P., Lenain, R., Thuilot, B., and Berducat, M.

(2002). Formation control algorithm for a ﬂeet of mo-

bile robots. 3rd International Conference on Machine

Control and Guidance.

Chan, E., Gilhead, P., Jelinek, P., and Krejci, P. (2011).

Sartre cooperative control of fully automated platoon

vehicles. 18th World Congress on Intelligent Trans-

port Systems.

Dafﬂon, B., Contet, J.-M., Gechter, F., and Gruer, P. (2012).

Toward a reactive agent based parking assistance sys-

tem. The 24rd IEEE International Conference on

Tools with Artiﬁcial Intelligence ICTAI, IEEE Com-

puter Society.

Daffon, B., Gechter, F., Contet, J., Abbas-Turki, A., and

Gruer, P. (2011). Intelligent crossroads for vehicle

http://safeplatoon.utbm.fr

platoons reconﬁguration. International Conference on

Adaptive and Intelligent Systems.

Daviet, P. and Parent, M. (1996). Longitudinal and lateral

servoing of vehicles in a platoon. IEEE Intel ligent

Vehicles Symposium, pages 41 – 46.

El-Zaher, M., Contet, J., Gruer, P., and Gechter, F. (2011a).

Towards a compositional veriﬁcation approach for

multi-agent systems : Application to platoon system.

First International workshop on Veriﬁcation and Val-

idation of multi-agent models for complex systems

(V2CS).

El-Zaher, M., Gechter, F., Gruer, P., and Hajjar, M. (2011b).

A new linear platoon model based on reactive multi-

agent systems. The 23rd IEEE International Confer-

ence on Tools with Artiﬁcial Intelligence ICTAI, IEEE

Computer Society.

Franck Gechter, Jean-Michel Contet, P. G. and Koukam, A.

(2010). A reactive agent based vehicle platoon algo-

rithm with integrated obstacle avoidance ability. Self-

Adaptive and Self-Organizing Systems.

Gechter, F., Contet, J., Gruer, P., and Koukam, A. Car driv-

ing assistance using organization measurement of re-

active multi-agent system. Procedia CS1(1): 317-325.

Gehrig, S. and Stein, F. (2001). Elastic bands to enhance

vehicle following. IEEE Conference on Intelligent

Transportation Systems, pages 597 – 602.

Hedrick, J., Tomizuka, M., and Varaiya, P. (1994). Control

issues in automated highway systems. Control Sys-

tems, IEEE, 14(6):21 –32.

Lamotte, O., Galland, S., Contet, J.-M., and Gechter, F.

(2010). Submicroscopic and physics simulation of

autonomous and intelligent vehicles in virtual reality.

Advances in System Simulation, International Confer-

ence on, 0:28–33.

Levine, W. and Athans., M. (1966). On the optimal error

regulation of a string of moving vehicles. IEEE Trans-

actions on Automatic Control.

o Yeong Yi, S. and Chong., K.-T. (2005). Impedance control

for a vehicle platoon system. Mechatronics.

Riess, R. (2000). Qualiﬁcation d’un tlmtre balayage laser

pour la robotique mobile : Intgration et exprimenta-

tions. Master’s thesis.

ICINCO2013-10thInternationalConferenceonInformaticsinControl,AutomationandRobotics