The SOCIAT Project: Simulation of Social Agents in Traffic

 

 

   To predict traffic behavior and provide a proper administration of traffic flow is a huge challenge. This includes maximization of roads utilization and prevention of traffic jams and related problems. We aim at building a simulator that considers drivers behavior and returns a forecast of the traffic situation in a reasonable time, using techniques of Artificial Intelligence (AI), especially Multi-Agent System (MAS), combined with game-theory in order to produce a realistic traffic model that reproduces drivers' behavior nearly to cognitive level.

   

  • Detailed Description:
  •    In every modern society the need for mobility is increasing daily and the capacities of the road network are satisfied or even exceeded. Thus the prediction and modelling of traffic flow is one of the future challenges to science. This project aims at building a tool for simulate, predict the state of the traffic and broadcast traffic messages to road users. However, we have to deal with the impact of traffic messages on drivers. Current systems are lacking an understanding of this reaction to messages. We use some well-known tools like the cellular automaton (CA) based microscopic traffic simulator (i.e. the simulation is at the level of driver-vehicle). The development of intelligent transportation systems (ITS) like dynamic route guidance systems affects the behaviour of drivers in a yet unknown way. Thus it is inevitable to model the cognitive and social aspects of the driver-vehicle entity more realistically. Artificial intelligence (AI) and, in particular, multi-agent systems (MAS) techniques open the possibility to model drivers in a more realistic way, at a level closer to the cognitive one. For instance, the behaviour of a driver can be described and explained via mental states like its beliefs and intentions (use of BDI logic), and forecasted via game-theoretic techniques. Thus, the objective of this interdisciplinary project is to use artificial intelligence and multi-agent tools in a microscopic simulation environment in order to treat drivers as intelligent agents.

    • Links:
    • Participants:
    • Finatial Support:
    • Brazilian National Research Council (CNPq)
    • German Aerospace Center (DLR)
    • Events:
    • Correlated Projects and Other Interesting Links:
    • Publications:
    • "SOCIAT: Simulation of Social Agents in Traffic"
      In: Proc. of the Fifth German-Brazilian Workshop on Informatics, Koenigswinter, 1999 (PS file available)
    • "Agents in Traffic Modelling - from Reactive to Social Behaviour"
      In: German Conf. on Artificial Intelligence - KI'99 (PS file available)

     

     

    Contact: Ana Bazzan