DSC'99 - Driving Simulation Conference July 7-8, 1999 - PARIS


Traffic Generation with the SCANeR II Simulator : towards a multi-agent Architecture

Alexis Champion
CS Communications & Systems - CSSI 3, rue Le Corbusier 94528 Rungis France

René Mandiau, Christophe Kolski
LAMIH &emdash; URA CNRS 1775 Université de Valenciennes et du Hainaut Cambrésis
Le Mont Houy - B.P. 311- 59304 Valenciennes Cedex France

Alexandre Heidet, Andras Kemeny
Renault, Direction de la Recherche 1, rue du Golf - API: TCR AVA 212 - F 78288 Guyancourt Cedex
tel: +33 (0)1 76 85 18 44 - fax: +33 (0)1 76 85 27 30 - e-mail: alexandre.heidet@renault.fr


Abstract

 Renault has been developing driving simulators for over 10 years. They are used for ergonomics and advanced engineering studies at Renault as well as for road traffic research, human factor studies and driver training in different laboratories and companies in Europe (espacially in France, the United Kingdom, Sweden and Norway).

These simulators use Silicon Graphics and/or PC image generation technology with up to 6 graphic channels and motion seat and/or platforms for kinaesthetic rendering for some of them. The real time simulation software, SCANeR II version 1.4, currently includes -among others- simulation session initialisation and monitoring, vehicle dynamics, traffic generation, visual and kinaesthetic modules.

The traffic generation software of SCANeR allows the user (i.e. the experimenter) to describe and initiate real time traffic with autonomous vehicles. Up to several hundreds vehicles can be rendered using a dynamic workload management system and several state machines standing for the different vehicles. The current software implementation has already been used for experimentations needing comprehensive scenarios or designed traffic situations.

One problem of this traditional approach is the difficulty to simulate a road traffic in critical and/or complex situations. Such situations mainly occur when several vehicles want to have access to the same road and/or when the number of involved vehicles is important. Yet, for many years now, researches mainly conducted in distributed artificial intelligence have allowed to study systems characterized by intelligent autonomous entities. These entities, or agents, must be able to coordinate their actions in order to reach their common goals. Such systems are named multi-agent systems.

Our work consists in modelling a road traffic as a multi-agent system. Vehicles have to cooperate and coordinate their actions in order to achieve a common aim (to reach and to drive on the same road) while avoiding endangering each other (accident risk).

In a driving simulator, agents (autonomous vehicles) must interact with each other and human beings (interactively driven vehicles). They constitute a traffic by intelligently anticipating, adapting and actively seeking ways to support one another. They must be able to communicate to create a cooperation and a coordination. Implementing a road traffic as a multi-agent system may improve the simulated traffic realism and contribute towards developing high level tools for traffic generation.

This paper will be composed of three parts. The first part will propose a survey of traffic simulation approaches. The second part will describe the current version of the traffic generated with the SCANeR II simulator. In the last part, technics under implementation according to multi-agent principles are discussed.