Optimization Of Driver Preview Course Decision-making Model Based On Particle Swarm Optimization

Making a comprehensive view of the status and application of driver behavior models research history , several direct control models had been established since 1950's. When driving the car, the driver always handles the vehicle based on the front road circumstance, because of the preview-follower driver model can sufficiently describing the characteristic of driver's preview, it is developed quickly. As it is mentioned, the typical driver models are the optimal preview control model brought forward by MacAdam based on the optimal control theory, the local optimal preview model brought forward by Reddy using local optimal algorithm and the driver preview optimal curvature model proposed by K.H.Guo in 1982 based on preview-follower theory and so on. These driver models can exactly simulate the characteristic of driver's controlling direction behavior when the driver control vehicle following the pre-given course. Because these driver models almost concern how does a driver preview and follow the pre-given course, in most simulation condition, the front road circumstance can be simplified a pre-given course, according to this pre-given course the driver can handles the vehicle.With the further research, researchers pay more and more attention to driver's direction control behavior under the real road circumstance. In the research of driver behavior model, driver models proposed by inland and overseas researchers mostly have the same precondition that vehicle preview course is usually been pre-given basically before the simulation, furthermore the driver dynamically control the direction to decide the vehicle course following the pre-given course, so the driver's decision process in the driver models doesn't roundly consider of the influence that the real road circumstance to the driver's deciding preview course. In the road circumstance, the road information identification is a road area, the driver dynamically decides vehicle preview course based on the road area.The driver's preview course decision-making process is a multi-objective fuzzy selection problem. The decision evaluation indexes are established according to the facts which influence the driver course decision, so we can choose the best course from the driver's preview courses. The driver preview course decision-making process is an optimization process, because the driver's inspect information changes from global to local, the existing driver models use grid search technique to search the optimal driver preview course. Its theory is: dividing the area estimated into gridding, in every gridding, we compute the objective function value and the restrict function value. To the points which are satisfied to the restrict condition, compare the value of the objective function, we choose point with the minimum objective function value to be the result in one iteration of optimization program. Then in the area which is around the point we have searched, divide the area into more little grid, and repeat the computation and comparison, stop the simulation unless the grid size is smaller than the given precision. This local search algorithms completely depend on the selection of the initial value and the design of the adjacent position function, if the design of adjacent position function or the initial value are not proper, the performance of the algorithms will be bad, furthermore it will lost the global optimization ability and get stuck at local optima easily.Particle swarm optimizers are a stochastic global optimization technique. The particle swarm algorithms find optimal regions of complex search spaces through the interaction of individuals in a population of particles. Compare every individual solution in search spaces as a particle without mass and solidity, these particles fly at a proper speed in search spaces, furthermore the particle dynamically adjust its speed according to the flying experience of all the particles, shortly, in every simulation iteration, the particle adjust its flying direction and speed according to its optimal value and the optimal value of all particles, this is a feedback structure program. Particle swarm optimization make the individual particle gradually to the better acceleration area according to the fuzzy evaluation indexes, finally search the optimal solution. Particle swarm optimization have"cognition"behavior and"social"behavior, because of the interaction between the partners, the particle adaptively adjust his direction and speed to the better area. Particle swarm optimization has better global optimization ability.The paper's major work is to develop the global optimization ability of driver preview course algorithm with the particle swarm optimization, and make the algorithm not overly depend on initial value and adjacent position function, improve the performance of algorithm.