Automatic Parking System Based On Model-free Adaptive Predictive Control

With the rapid development of artificial intelligence technology and control technology,driverless technology continues to make new advances and breakthroughs.As an important part of the Intelligent Driving Assistance System,the Automatic Parking System(APS)plays a very important role in reducing parking accidents and improving parking safety.The Automatic Parking System consists of three functions: parking space detection,path planning and path tracking.Most traditional path planning methods ignore vehicle steering constraints,resulting in a discontinuous path curvature that causes the vehicle to spin in place and is prone to wear and tear on the vehicle.In addition,most of the existing automatic parking control methods are model-based and have poor portability.Therefore,it is of great theoretical significance and practical application to study path planning methods that satisfy vehicle constraints,as well as model-independent and more portable automatic parking control methods.In response to these problems,this thesis proposes a new solution for automatic parking system-ultrasonic distance-based parking space detection scheme,a B-spline curve-based path planning scheme,and a model-free adaptive predictive control(MFAPC)based path tracking control scheme.A comparative simulation study of MFAPC,Model Free Adaptive Control(MFAC)and conventional PID methods is carried out via the Car Sim/Simulink joint simulation platform.The main work is as follows.(1)A parking space detection scheme based on ultrasonic distance measurement is designed.The basic ranging principle and main characteristics of ultrasonic radar are described,and the corresponding parallel parking space detection logic algorithm is given for parallel parking conditions.The feasibility of the designed parking space detection scheme is verified by simulation.(2)The kinematic model characteristics of the driverless vehicle are analyzed,and the vehicle and environmental constraints are combined to plan the automatic parking path.The characteristics of three algorithms based on B-spline curve,circular linear method and double circular arc method are investigated,and the superiority of the B spline curve path planning scheme is verified by conducting simulation comparison experiments through MATLAB.(3)According to the control objective of the automatic parking system,the modelfree adaptive predictive control based parking path tracking control scheme is given under the consideration of constant vehicle speed.The automated parking system is equivalently transformed into a full-format dynamic linearized data model,which in turn gives modelfree adaptive predictive control algorithms,pseudo-gradient estimation algorithms and prediction algorithms and reset algorithms based on optimization metrics.The simulation of the complete automatic parking process is given in the built Car Sim/Simulink joint simulation platform,and the MFAPC scheme proposed in this thesis is compared with the MFAC scheme and the PID scheme for the simulation study.The results show that the MFAPC-based automatic parking system can achieve the complete automatic parking process,and the MFAPC scheme has higher accuracy in lateral error tracking compared with the MFAC scheme and the PID scheme.There are 45 figures,9 tables,61 references.