Agent-based Modeling Of Traffic System

Research on traffic flow modeling is of both theoretical significance and realistic value on simulating the real traffic phenomenon, solving traffic problems and predicting potential traffic conditions. However, the existing modeling methods, which describe the whole traffic system from the macro view or describe the particular aspect of drivers from the micro view, are unable to reflect the heterogeneity, dynamic and fuzziness of drivers and their interactions in traffic flow. Therefore, this article proposed an agent based modeling to simulate the real traffic flow.Firstly, two typical scenarios are analyzed to summarize the general modeling requirements in drivers’ behavior. Then, based on the composition of fuzzy information processes and neural network, the framework of drivers’ behavior is constructed, which include four parts of recognition, decision, action and emotional accumulation. According to the differences of drivers’ action in different environments, free flow, congestion flow and crossing flow are introduced to describe specific changing of behavior in these conditions. And then, agent based modeling is utilized, which including the construction of traffic road, generating vehicles, forming OD matrixes and defining agents’ attributes and behavior rules. The modeling of interactions between agents and between agent and environment make whole traffic flow system come into being.The comparing of experiment with real measured dates and classical modeling shows that the modeling in this article can better illustrate the real traffic flows. According to the experiment, it shows that the better the driver behaviors, the less likely he against the rule and the longer he takes to accumulate his emotion, and the amount of irregular behaviors in free flow is smaller than in congestion flow and larger than in crossing flow, and standardizing drivers’ behavior can largely improve irregular phenomenon in traffic flows.