Research On Traffic Flow Models Based On Cellular Automaton And Fuzzy Inference

Traffic flow model has been studied to improve the control, management and organization of traffic system since a long time ago. After the Kinematics Model, Car-following Model and Gas-Kinetics-based Model, cellular automaton (CA) model has been widely applied in study of traffic flow in recent years. CA model is a useful tool for simulation of nonlinear complex systems. NaSch model is the most popular cellular automaton model for traffic flow. On the basis of NaSch model, some modified models were proposed. These above traffic flow models modify the update rules or adjust randomization deceleration probability according to factors about drivers" features, neighbor vehicle and vehicles’ speeds. But in those modified models, the car ahead is stationary and the ambiguity of influencing factors is not taken into account.In this paper, a new traffic flow model based on spatial headway and velocity difference is proposed. The randomization deceleration probability depends not only on the positions of this car and the car ahead of it, but also on the velocities of the two cars. The car ahead is dynamic for the following car, not stationary. The computer simulation results show that the proposed model is able to reproduce some of the real macroscopic traffic characteristics, and the velocity distributions for the new model are relatively reasonable.To make the traffic flow model based on cellular automaton more close to reality, fuzzy logic system is designed and integrated into the traffic flow model in this paper. In the proposed model, spatial headway and velocity difference are considered as two main driving factors. Randomization parameter of every vehicle changes according to the above two factors at every time step. Spatial headway and velocity difference are the inputs of fuzzy logic system, the output is the randomization deceleration probability of the current vehicle at the next time step. The computer simulation results show that the proposed model is able to reproduce some of the real macroscopic traffic characteristics such as stop-and-go waves, hysteresis phenomenon. At a certain density, the calculated randomization parameters are basically consistent with the actual traffic situation.