Research On Car-Following And Lane-Changing Model Based On Driver's Information Processing Characteristics

Microscopic traffic simulation system is an effective tool for evaluating traffic management measures and advanced technologies. Car-following model and discretionary lane change model are two main models of Microscopic traffic simulation system. Well comprehending the characteristics of driver is the fundamental of modeling driver behaviors. Inaccuracies and anticipation and the non-linear relations between input information and driver's reactions are main characteristics of driver. It is very difficult to model driver's information processing behaviors by traditional mathematic methods. Artificial intelligence is a good method to model these characteristics of driver. This thesis developed driver's anticipation model using fuzzy qualitative simulation, developed car-following models by fuzzy logic and artificial neural network, and developed discretionary lane changing model using fuzzy logic. This thesis included five parts.The first part developed an anticipation model of driver in car following process. 38591136Traditional car following models mainly concern the reaction results of drivers but not the reaction processes of drivers. Then the traditional car following model does not modeling driver's anticipation behaviors. However, analyzing traffic management measures and advanced technologies need to know the effects of every phase of driving. Developing driver anticipation model can help analyzing driving behaviors and traffic management measures and advanced technologies. Fuzzy qualitative simulation can simulate the ability that human process qualitative information. This thesis developed driver's information perception and anticipation model using fuzzy qualitative simulation. This part introduced the theory, conceptions and arithmetic of fuzzy qualitative simulation. This part introduced the variables selection, fuzzy space, and rules in driver's anticipation model based on fuzzy qualitative simulation. The results showed that this model can simulate the ability of driver process inaccuracy information and anticipation ability. This model can also explain the reason that time headway is shorter than driver reaction time.The second part developed a car following model based on fuzzy logic .The information that driver percept is inaccurately. The information used by driver decide process include subjective and impersonality information. Fuzzy logic is a good way to process subjective and objective information. So the car following model can be developed by fuzzy logic. The decision objective is variable under different traffic state. So this thesis developed two models for car following process according to the different decision. This part provided the conditions for different decision. This model can simulate real traffic phenomenon.The third part developed a discretionary lane change model based on fuzzy logic. This thesis introduced driver's aggressive as lane change desire parameter. So this model considered the subjective and objective reasons for discretionary lane change in the same time. It also considered the time accumulative lane change desire by introduced the accumulative available acceleration as a evaluation parameter.The fourth part developed an Artificial Neural Network (ANN) based car following model. The input and output of driver in car following process is non-linear. The details of ANN-based car-following model are provided in this part. This part also gives an example to demonstrate the practical application of the model. The results show that ANN is an effective tool to simulate driver's decision behavior.This part developed the ANN training algorithm based on Particle Swarm optimization.It can avoide local Optimization.The fifth part developed the traffic simulation system to realizing car following and discretionary lane change model designed by this thesis. This part introduced the detail design of driver information anticipation model, car following model and discretionary lane change model.In the end, this thesis summarizes the findings and achievements, and brings forward the issues for further research.