Research On Hierarchical Planning And Motion Control For Automated Valet Parking In The Park Scenario

The contradiction between the continuous rise of car ownership and the insufficient construction of urban parking lots has caused serious parking problems.Problems such as insufficient number of parking spaces and small parking space have seriously hindered the completion of vehicle travel tasks.Automated valet parking technology is an autonomous driving technology for the park scenario,which can realize fully autonomous driving in the park and effectively solve the vehicle travel problem.This paper relies on the major science and technology special project of Jilin Province-Automated valet parking development and application research（No.20200501009GX）,and takes automated valet parking as the research object to investigate the global path planning,obstacle avoidance path planning,parking path planning and lateral motion control methods of autonomous vehicles in the park scenario.The main contents of this paper are as follows.（1）The edge detection algorithm is introduced into the processing of environmental maps to improve the processing of raster maps when performing path planning for driving in the park.Based on the idea of hierarchical planning,the global path planning method using Hybrid A* algorithm is investigated by ignoring the raster map of dynamic obstacles in the park.Considering the situation of mixed flow of people and vehicles in the park and the complex driving environment,the obstacle avoidance planning problem of the park is modeled by numerical optimization method to study how to obtain feasible obstacle avoidance paths considering vehicle obstacle avoidance constraints and motion constraints.（2）In view of the typical parking space form and parking conditions in the park,this paper studies how to establish a parking planning framework that adapts to a variety of typical parking scenarios in the park.By analyzing the typical parking spaces in the park,considering the vehicle obstacle avoidance constraints and motion constraints,the paths of single-step parking and multi-step parking in the park are derived based on the combination of arc-linear approach.The multiple parking methods are combined to establish a relatively complete parking planning framework.Finally,the paths are processed into trajectory sequences containing velocity information for controller tracking driving.（3）After obtaining the automated valet parking trajectory of the park through the planning algorithm,this paper studies how to design the lateral controller to make the vehicle accurately track the desired trajectory.The lateral motion controller of MPC is established based on the prediction model of the lateral displacement of the pre-sighting point.By making full use of the future path information,the desired transverse angular velocity and the lateral displacement of the pre-sighting point are taken into account in the design of the lateral motion controller of MPC,so that the prediction capability of MPC can be improved without increasing the prediction time domain.（4）In this paper,we build a joint simulation platform using Prescan,Carsim and Matlab/Simulink software,establish an automated valet parking simulation scenario based on Prescan,combine the vehicle model in Carsim,build the parking path planning algorithm,path processing algorithm and lateral control algorithm in Simulink,and integrate and verify the above algorithms.Finally,the proposed planning and control method for automated valet parking is verified in real vehicles with the self-developed autonomous driving test platform.