Research On Automatic Parking Strategy For Complex Obstacle Scenes

With the rapid growth of car ownership,there is a huge gap in parking infrastructure,and the further narrowing of parking spaces and the continuous complexity of the parking environment have made the problem of "parking difficulty" increasingly prominent.The parking accident rate,parking time,and driving technical requirements have continuously increased.As an important way to effectively solve the above problems,automatic parking systems have broad research significance and market prospects.This paper studies the automatic parking system,to achieve parking space and obstacle perception,multi strategy dual optimal parking path planning,and tracking control in complex parking scenarios.The main research is as follows:First of all,the camera and ultrasonic radar are fused to realize parking space recognition and obstacle perception.The camera uses Canny edge detection,Hough line recognition and screening clustering to realize the recognition of the parking space line,and the ultrasonic radar analyzes the jumping characteristics of the distance data to identify the parking spaces without parking space line or the parking space line is blocked,and realizes the comprehensive parking space recognition through sensor fusion.At the same time,more comprehensive and accurate obstacle information is obtained through the integration of YOLOv4 perception and ultrasonic radar ranging through the camera.Through the experimental verification of parking space recognition and obstacle perception in real vehicle environment,the results show that this method can achieve comprehensive and accurate parking space recognition and obstacle perception.Secondly,based on geometric method,multi-strategy parking path planning is carried out for narrow parking spaces with arbitrary vehicle starting positions and postures.The Ackerman steering model is established and the parking constraints are analyzed,and the parking longitudinal speed is planned.Introducing the transition curve,and according to the arc-line-transition curve method,three parking strategies are put forward: one-time parking strategy,one-time parking strategy after forward adjustment,and multiple-time parking strategies.For the narrow vertical parking space and horizontal parking space,the parking path is planned with any initial vehicle position.Simulation is carried out in the Matlab environment to verify the effectiveness and applicability of the path planning method.Then,considering the impact of complex obstacles,based on multiple parking strategies and obstacle avoidance strategies,the dual-optimal real-time path planning in multiple parking spaces environment is realized.Analyze the characteristics of obstacles,consider the interaction between obstacles,use LSTM to predict the future trajectory of obstacles,summarize the collision conditions of obstacles,and propose three obstacle avoidance strategies: parking and waiting,changing parking strategy and vehicle detour.In the multi-slot environment,the comprehensive dual-optimal evaluation index of parking path and target parking space is proposed,the trigger mechanism of path re-planning is introduced considering the random characteristics of obstacles,and the mechanism of pre-screening and logic judgment simplification is introduced considering the computational efficiency,to realize real-time dual-optimal parking path planning.The simulation results in Matlab environment show that the path planning method for complex parking scenes is effective,robust and applicable.Finally,NMPC is used horizontally and double PID are used vertically to realize tracking control.The vehicle lateral control is realized based on nonlinear model predictive control,and the vehicle longitudinal control is realized based on position PID and speed PID.The simulation results in the joint simulation environment of Matlab/Simulink and Car Sim show the effectiveness of the control method in this paper.In addition,the effectiveness and applicability of the parking system in this paper are further verified by the real vehicle experiment on automatic parking path planning and tracking control.