Research On Automatic Parking Technology Based On Self-organizing Fuzzy Neural Network

Since the second industrial revolution,the car enters people's lives gradually.Now the car has entered thousands of families and become the main means of transportation.At the same time,with the increasing population of vehicle,city traffic is jammed and the parking problem has become more prominent.Parking space has become increasingly narrow and the actual environment of parking has become more and more complex.The driver needs to complete the tedious process of parking in a limited space,which is low efficient and safety can't be guaranteed.Especially to the driver who lack of experience,it is difficult to complete the consistent parking process.People's negligence can easily lead to accidents.Therefore,the development of automatic parking system is necessary to complete the tedious parking process.At present,fuzzy control is adopted by most of the parking systems,control rules of the existing fuzzy control are usually based on expert experience,which is subjective.When the problem is complex,fuzzy rules are often very difficult to access.It's easy to affect the system time decision if rules are complex,or rules are too few to keep system control accuracy and stability.In this paper,the self-organizing neural network is used to determine the fuzzy rules.The determination of membership function is combined with the neural network,and the corresponding parameters are adjusted by the self-learning and adaptive ability of the neural network.Finally,the fuzzy neural network structure is determined.The movement of the car in the actual environment is often affected by multiple factors.This paper established a simplified kinematic model of the car and the relations between the radius,steering wheel and speed were discussed,and MATLAB simulation platform was used to study the development process of parallel parking.Parking space coordinates,sampling time,body size and etc.were set.Fuzzy controller and fuzzy neural network controller were used in the car parking process.The controller,based on feedback of body angle,front wheel steering and coordinates,controlled the car movement to complete parking simulation.A physical model of car was set up to verify the algorithm above.The location information was obtained from the ultrasonic sensors,which was used to draw parking map and identify parking spaces automatically.The trained fuzzy neural network was transplanted to the ARM processor to control the movement of the model car.At the same time,the location information was obtained to correct the movement in time and provided anti-collision warning and the car backed into the garage safe and smoothly.