The Study Of Urban Traffic Control System Based On Agent

Urban Traffic Control System is one of important ways to improve the operational efficiency of urban transport and it is important sign of intelligence and modernization of urban transport. The advanced urban traffic signal control system has an important effect on improving traffic conditions and sufficiently utilizing roadway capacity of existing road. Agent technology is the current research topics on Distributed Artificial Intelligence, transportation system is a distributed, nonlinear, time-varying stochastic systems, the paper applied Agent technology to urban traffic control system.Firstly, a new frame of urban traffic control system was put forward on the basis of introducing the actuality and control method of urban traffic control system in our cities, and summarizing Multi-Agent technology. The system can be divided into three-layer: the crossroads of Agent, the promoter region of Agent, the central Agent. The system hands the authority of deciding on control strategy to intersection. According to acquired real-time traffic information, the intersection controls traffic signal through cooperation. The intersection improves capability of intellect strategy. So control system can adapt flexibly to complex features of traffic system. It supplies a probability for real-time intellect control.Secondly, on the basis of putting forward urban traffic control system, the paper researched on urban control coordination algorithm based on Nash equilibrium and relevant theory of dynamics division of traffic sub-area. In addition, the paper detailed control strategy of the entire system and sub-zone control strategies. Game theory was applied to urban traffic control system, the research realize coordination between the intersection of Agent.Finally, the paper researched on precondition and key technology of urban traffic control system coordination -Agent Short-term traffic forecasts. Using genetic algorithm optimization of the RBF neural network to predicted short-term traffic to prove that the GA-RBF neural network to predict the effectiveness and feasibility.