Design And Research Of Automatic Parking System Based On Model Predictive Control

With the continuous increase of car ownership,the number of parking spaces in cities becomes more and more tense.The traffic accident problem caused by parking is very serious.Automatic Parking System is an important component of Advanced Driving Assistance System.It can control the vehicle moving into the parking space quickly and safely without manual intervention,And it effectively reduce the burden of drivers and the probability of accidents during parking.According to the current research status of automatic parking system,in this paper,we mainly study the curvature continuous path planning method,and applying model prediction control method to path tracking controller.First of all,this paper adopts the combination of ultrasonic sensor and wheel speed pulse sensor to realize the detection of parking space and to obtain the initial position of parking.In order to improve the stability and accuracy of measurement data and prevent the failure of a single radar,it is proposed to use the two ultrasonic radars on the same side to detect the parking space at the same time,and combine the similaritybased data fusion method to obtain more accurate parking space information.Then,the low-speed motion process of the vehicle is studied,and the kinematics model based on the center of the rear axle is established.The motion law of the reference point under the input of the vehicle speed and steering wheel angle is clarified,and the law is extended to the vertices of the vehicle body.The kinematics and collision constraints of single-step parallel parking are analyzed and used as nonlinear constraints.The path optimization function is designed based on B-spline curve theory.On this basis,multiple parking starting points are selected for MATLAB path planning simulation,which verifies the rationality of the path planning method.In order to track the planned target path,firstly,the EKF-based track estimation method is used to filter out the noise signal in the sensor to obtain accurate vehicle local positioning information.Based on the deviation between the vehicle's real-time pose and the target path,a path tracking controller based on model predictive control is designed.The appropriate objective function is selected to convert the tracking control problem into a convex optimization quadratic optimal solution problem.The problem of parameter selection is studied.At the same time,a pure persuit control method widely used for path tracking is introduced,which is used as a comparison verification of controller control effects.Through the joint simulation of MATLAB/Simulink and Carsim,the position error and path angle error of path tracking under the two control algorithms of model predictive control and pure persuit control are compared.The results show that the control effect based on model predictive control is better.Finally,the effectiveness of the parking space detection algorithm based on dual ultrasound radar data fusion is verified on the real vehicle.The function verification of the parking system control strategy is carried out,and the error analysis is carried out in combination with the CAN bus signal in the vehicle,which proves the rationality of the planning path and the effectiveness of the path tracking controller.