Research On Path Planning And Tracking Algorithms Of Automatic Parking System

With the continuous increase in car ownership,the difficulty of parking has gradually become a major problem for people to drive.The automatic parking function realized by intelligent driving technology can effectively reduce the burden of the driver and reduce the number of accidents in the parking process,so it has broad market prospects and practical significance.The automatic parking system must not only consider the adaptability to different parking environments,but also meet the consumer experience.Path planning and motion control are the key components of the automatic parking system,and their performance determines the control accuracy of the automatic parking system.Aiming at the problems of small scope of application of existing path planning algorithms,as well as the problems of low control accuracy in motion control,this paper optimizes and improves the planning and control algorithms.The specific research content is as follows:(1)A vehicle kinematics model is established based on the Ackerman steering principle,and the motion equation of the vehicle profile is derived based on the simplified vehicle model.The vehicle model is built in Simulink and compared with the Car Sim vehicle model.The analysis results show that the simplified vehicle model established in this paper can accurately reflect the vehicle’s motion state at low speeds,which verifies the feasibility of the model.(2)In response to the specific requirements of the automatic parking system for path planning,the traditional rapidly-exploring random tree(RRT)algorithm is improved.Firstly,the principle and flow of RRT algorithm,sampling method and node expansion method are analyzed.The advantages and disadvantages of the RRT algorithms are obtained.Aiming at the problem that the original algorithm path point connection does not meet the vehicle constraints,the Reeds-Shepp curve is added to expand the nodes.To improve the expansion speed of the original algorithm,a target bias strategy is added during sampling.Obstacle avoidance sampling and constrained sampling are proposed to improve the quality of sampling nodes of the algorithm and search efficiency.Finally,the redundant points are trimmed on the path generated by RRT,and the B-spline curve is used to smooth it.The simulation results show that the improved path planning algorithm can obtain a parking path that meets safety,feasibility and comfort.(3)In order to track the parking path generated by the path planning,the longitudinal and lateral motion control algorithms in the parking environment are studied.For longitudinal control,a vehicle speed controller is designed based on PID.For lateral control,the wheel steering angle is controlled based on model predictive control algorithm,and vehicle dynamics and safety constraints are set.According to the characteristics of parking conditions,the cost function and weight matrix are optimized.The simulation is carried out in Simulink,and the results show that the controller can accurately control the vehicle to follow the planned path.(4)The parking environment is designed according to industry standards,and a joint simulation platform is built based on Simulink and Car Sim.The improved automatic parking system algorithm in this paper is simulated and verified in several different scenarios.The simulation results show that the automatic parking system algorithm proposed in this paper can plan feasible paths in various scenarios,and can control the vehicle to successfully park in the parking space.The lateral deviation is about 0.1 meters,and the vehicle heading angle deviation does not exceed 3° after parking.