Design And Experimental Research On Path Planning Algorithm For An Autonomous Vehicle

With the advancement of computer technology and information technology,vehicles are gradually developing in the direction of intel igence and unmanned.Path planning algorithm is one of the important foundations for intel igentization of autonomous vehicles.Therefore,many companies and universities at domestic and foreign have conducted a lot of research on the path planning algorithm of autonomous vehicles to meet the needs of driving in complex traffic environments.In this thesis,based on the performance indicators of autonomous vehicles,vehicle kinematic constraints and computational efficiency,a series of researches on the path planning algorithm of autonomous vehicles are carried out.The global reference path planning algorithm for autonomous vehicles is designed,and the risk field model and dynamic local path planning algorithm are established.Relying on the real vehicle platform provided by an autonomous driving company,the relevant content studied in this paper is applied to the actual development of autonomous vehicles,and has achieved certain results.Main research as follows:First,the research status of autonomous vehicles at native and abroad is investigated.The typical path planning algorithms are listed and the characteristics of each algorithm are analyzed.Secondly,the problem of the global path planning algorithm is slow and cannot balance the vehicle non-integrity constraints and the vehicle performance indicators.In this paper,a global reference path planning algorithm based on Batch Information Tree is designed.The distance transformation map and k-d tree data structure are introduced.On the basis of not generating a large number of operations,the vehicle kinematics constraints are satisfied and the driving safety,efficiency and stability are considered when generating a global reference path.Thirdly,aiming at the problem that the driving traffic environment description dimension is large and the accurate model is difficult in the partial path planning of the autonomous vehicle,the driving traffic environment is abstracted into the driving risk field model.According to the data obtained by the sensors,the obstacle risk distribution model and the risk assessment function are designed.The obstacles are described as dynamic risk fields and static risk fields,respectively,according to the dynamic and static properties of the obstacles.It greatly reduces the description dimension of the traffic environment and provides a reference for the local path planning algorithm.Fourthly,the local path planning problem for autonomous vehicles with horizontal and vertical coupling is difficult to solve,and the problem of real-time requirements cannot be met.The local path planning algorithm for autonomous vehicles is designed.In this paper,the local path planning problem is introduced into the Frenet coordinate system for decoupling and solving the motion candidate set.The horizontal and vertical motion candidate sets are then coupled again,the candidate motions that do not satisfy the dynamic constraints are deleted,and the local reference path is preferentially output according to the loss function.Finally,the variable frequency rolling optimization strategy is used to realize the dynamic local path planning of the autonomous vehicle based on the real-time requirements.Fifthly,based on a real vehicle platform provided by a company,the dynamic local path planning algorithm studied in this paper is verified in real under various typical conditions and the results are analyzed.At the same time,the computational efficiency of other types of RRT global reference path planning algorithm and BIT(Batch Information Tree)global reference path planning algorithm under different size maps are compared.