Research Of Modeling And Applications On Agent-Based Traffic System

Modeling of traffic flow is of great significance on analyzing traffic phenomenon, predicting the effect of traffic management measures, planning road network and solving traffic problems. Microscopic traffic flow has unique superiority in depicting detailed and dynamic characteristics over macroscopic and mesoscopic ones, among which, agent-based modeling can reflect the differences and interaction of the individual behaviors compared with cellular automata modeling. In view of the neglect of heterogeneity and autonomy of drivers and fuzziness and randomness of traffic information, the article proposes an agent-based modeling of drivers’ behavior and traffic flow distribution to simulate the real traffic flow, and resolves the problem of road network planning and police deployment.Firstly, several typical scenarios are analyzed to describe the changes of drivers’ psychology, emotion and behavior in complicated traffic environment. The drivers’ behavior model is constructed based on agent controlling the driving patterns switching among free driving, following, lane changing and turning. And then, the traffic flow distribution analyzes the fuzzy stochastic impedance and considers the effect of neighbor drivers, in which the traffic flow is influenced by the external road environment and internal drivers’ characteristic. Eventually, the paper applies the model proposed above to the road network programming and police deployment with the actual traffic consideration. The comparing of experiment of drivers’ behavior model with real measured dates shows that the car following mode, lane changing model and intersection steering model in this article can better illustrate the real traffic flow. The comparing of experiment of traffic flow distribution model with real measured dates and classical model shows that the present model has higher precision and can better reflect the randomness of route selection and time-variability of Origin-Destination matrix. The experiment results and application demonstrate that the modeling of agent-based traffic system exposing the heterogeneity of drivers and the complexity of traffic environment is an effective and realistic method to solve the traffic problem.