Research On Behavior Decision Making And Trajectory Planning Algorithms Of Driverless Vehicle In Park Environment

A driverless vehicle is an intelligent vehicle that automatically avoids obstacle and drives safely to the target without driver involvement.How vehicles choose their behavior and drive stably and safely in the environment have received more attention.Due to the campus environment with low vehicle speed and uncomplicated traffic conditions,the problem of behavior decision making and trajectory planning for driverless vehicles in the campus environment is investigated.This thesis has already studied the following.(1)To study Q-Learning driverless vehicle behavior decision system based on dynamic search strategy.The Q-Learning algorithm is combined with the dynamic search strategy to solve the problem that the Q-Learning algorithm is prone to fail to learn the optimal behavior and convergence difficulties.The DSS-Q-Learning algorithm is verified to be suitable for solving the behavior decision problem of driverless vehicles in campus environment through comparative experiments.(2)To study the global path planning method of driverless vehicle based on Multi Cost-Floyd(MC-Floyd).Considering the road information and other factors in the scene,the minimum number of turns of the vehicle is considered on the basis of the shortest path,so that the global path planned by applying Floyd algorithm is the optimal path for the vehicle driving in the actual scene.Based on the map node information of the real campus environment,the comparison experiment verifies that MC-Floyd can plan an optimal path for the driverless vehicle that meets the conditions.(3)The Potential Energy Reconstruction-Artificial Potential Field(PER-APF)algorithm is proposed for local lane change trajectory planning of driverless vehicle.By constructing a virtual region and reconstructing the physical potential field,the problem of local optimum and unreachable target easily occur by applying APF algorithm.Different simulation scenarios are set up and compared with the APF algorithm to verify the feasibility and effectiveness of the PER-APF algorithm in solving the local lane change trajectory planning problem for driverless vehicles.(4)Based on the driverless experimental platform,the MC-Floyd based global path planning algorithm and the PER-APF based local lane-change trajectory planning algorithm are verified.The experimental results show that the MC-Floyd algorithm and the PER-APF algorithm proposed in this thesis have the value of application in a practical environment.