Research On Path Planning And Tracking Algorithm Based On Reachable Space Of Automatic Parking

As a kind of important and efficient transportation tool in daily life and work,car has become an indispensable part of our lives.However,with the annual increase in the number of cars in our country,parking a car into a small,complex,and harsh parking space is becoming more and more difficult to the drivers.With the development of the automotive industry in the direction of intelligence continuously driven by current technological progress,cars with automatic driving and automatic parking functions are favored by consumers.A car equipped with an automatic parking system can autonomously complete the automatic parking process in a small,complex,and harsh parking space,which solves the parking troubles the majority of drivers facing with.In order to solve the current problems in the development and research of automatic parking at home and abroad,this article takes "environment sensing is the premise,path planning is the foundation,and path tracking is the key" as the principle,and algorithms are carried out for the three modules of environment sensing,path planning and path tracking.The main research contents are as follows:Firstly,by analyzing the working conditions of the parking process and determining the type of automatic parking,the kinematics model of automatic parking is established according to the vehicle motion characteristics.The parking kinematics model is the theoretical basis for subsequent algorithm research and experimental verification of automatic parking environment perception,path planning and tracking.Secondly,the parking space detected by the ultrasonic sensor has a large error,this paper proposes the automatic parking reachable space algorithm consisting of SVM automatic labeling and classification and LT-IEPF line segment fitting,which can effectively classify the ultrasonic point cloud by the support vector machine classifier and segment and fit the point cloud by the fitting algorithm,so that the outline of the target parking slot and the target pose are determined.The results show that the classification accuracy of the ultrasonic point cloud is 92.47%,and the lateral reachable space of the selected six parking spaces is increased by more than 0.6m,which has a good effect.Thirdly,in order to solve the problem of low path planning success rate caused by factors such as diverse parking poses and complicated path interference,this paper proposes a feasible path detection algorithm to detect whether the parking path is feasible;Combined with the detection algorithm,a parking path planning algorithm composed of one-segment,two-segment,three-segment and multi-segment parking paths is given.Finally,for the designed 18 parking scenarios test cases,the algorithm in this paper can successfully generate feasible parking paths.Finally,the lack of robustness of the path tracking algorithm is an important reason why the vehicle cannot be parked accurately and effectively.Therefore,this paper proposes a parking path tracking algorithm based on sliding mode control to determine the vehicle’s front steering angle δ control law.Build a MATLAB/Simulink path tracking simulation platform,and conduct path tracking comparison experiments and a single experiment for the parking paths obtained in different parking scenarios.According to the analysis of the automatic parking path tracking distance error,heading angle error and front wheel angle of each scene,the feasibility and robustness of the path tracking algorithm in this paper are verified.