Research On Automatic Parking Path Planning And Tracking Control Based On Panoramic Image

With the improvement of people’s living standards,the number of cars in our country is also increasing rapidly,resulting in lacking parking slots and relatively smaller parking slots in cities.Subsequently,the issue of " parking difficulties" has emerged.As a part of the unmanned driving,the automatic parking system has a very broad market demand.This paper uses the vehicle’s panoramic surround view system to propose an automatic parking path planning and tracking control method based on panoramic images.The specific research contents are as follows:It introduces the parking slot detection method based on panoramic images used in this article: using a multi-target detection network to determine the head and landmark points of the parking slot,determines the complete parking slot through geometric relations and the parking slot location according to the ratio of pixels to actual dimensions.The vehicle model is simplified according to the Ackerman steering geometry,and the low-speed vehicle kinematics model is derived,which provides convenience for subsequent research.According to the three types of parking slots obtained by the parking slot detection,the traditional method of connecting circular arc s and straight lines are used to plan the parking path.For vertical parking slots and oblique parking slots,according to the difference in the lateral distance between the starting position of the vehicle and the parking slot entrance line,a single-turn and double-turn parking path planning algorithm is proposed.For parallel parking slots,reverse inference is used to calculate the minimum parking slot size required for two-step parking.For parking slots larger than this size,two-step parking path planning is carried out,and for parking slots smaller than this size,four-steps methods is taken.According to different lateral distances,it can also be divided into two methods.Using Matlab to establish a path planning simulation model,and verifying the above planning methods.The results show that they are feasible.Using the obtained low-speed kinematics model and combining it with the principle of model predictive control(MPC),to design the automatic parking path tracking controller.The algorithm is derived in the article.The soft constraint method is selected to set the objective function and constrain the control quantity and the control quantity increment,then calculating the control value of each sampl ing period.A joint simulation model of Carsim and Simulink is established on the computer to verify the designed path tracking controller.The results show that it has practical significance.Four fish-eye cameras are used to build a panoramic surround vi ew system on the Mammoth IV chassis.The MINI industrial computer on the chassis is pre-installed with the Ubuntu system,which is used as the software system platform,equipped with the Robot Operating System(ROS)as the software architecture,communicat ing through the CAN communication module and the underlying control system to construct an automatic parking system.Three types of parking slots are designed according to the size of the vehicle.The results of many real-vehicle experiments show that the automatic parking system built in this paper can be applied to ordinary parking scenarios.It has been verified that the automatic parking path planning methods proposed in this paper have good implementability and can reduce the accuracy requirements for tracking control.The established automatic parking system can accurately identify the parking slots and safely complete the parking operation,and can minimize the requirements on the size of the parking slots when the parking space permits.