Research On Autonomous Vehicle Path Planning Methods Based On RRT

The rapid development of artificial intelligence technology has promoted another revolution in traditional industries.Autonomous vehicles,as a typical represent of this artificial intelligence revolution,are in the process of rapid development.As the core technology of autonomous driving,path planning technology faces an unprecedented challenge.Existing automatic driving path planning methods generally have problems such as considering vehicles into particles,the path does not conform to the vehicle kinematics constraints,low planning efficiency,and poor safety.Therefore,this thesis conducts methods research on the path planning of autonomous vehicles.The main contents of this thesis are as follows:Firstly,to solve the problems of traditional path planning considers vehicles into particles,the path does not meet the vehicle kinematics constraints,and the low planning efficiency,an AV-RRT(Autonomous Vehicles based Rapidly Exploring Random Tree)non-particle path planning method for autonomous vehicles is proposed.Before path planning,the obstacle expansion process is performed on the environmental space to ensure that the distance between the obstacles on both sides of the planned path is not less than the width of the vehicle,and the vehicle can pass safely when driving in a straight line.In the random sampling stage of the algorithm,the maximum turning angle of the vehicle is used to limit the angle between paths,and the direction of the sampling points that do not meet the maximum turning angle of the vehicle is corrected to ensure that the planned path meets the requirements of vehicle kinematics constraints.The algorithm also uses dynamic step size instead of traditional fixed step size for random tree growth,and uses the strategy of "fast growth,steady rise,fast recovery" for random tree branch growth,which shortens path planning time and improves algorithm efficiency.Secondly,the autonomous vehicles path planning methods still have collision risks during vehicles turning.A NPCD(Non-Particle Collision Detecting)collision detection algorithm and PM(Path-Modification)path in the vehicle turning process are proposed to ensure the safety of vehicles driving along the global path.Geometric constraints are used to represent the safety threat area that may have collision risks formed by the vehicle’s wheel traces and obstacle positions during the vehicle turning process,and accurately detect the safety risks faced by the vehicle when driving along the path.The shortest distance from the safety cavity is the standard design path modification strategy to avoid the possible collision risks of the vehicles.Finally,the validity and feasibility of the method proposed in this thesis are verified through comparative simulation experiments.By comparing with the traditional RRT algorithm and the KB-RRT algorithm,it is verified that the path planned by the AV-RRT algorithm meets the functional effectiveness of the vehicle kinematics constraints,as well as the superiority in performance indicators such as the time consumption of path planning and the number of path nodes.At the same time,it is verified the functional effectiveness of collision detection algorithms and path modification strategies.