Automatic Parking System Research Based On Multi-sensor Information Fusion

At present,the automotive intelligent technology is developing rapidly.Automatic parking system as an important part of intelligent cars,has received the attention of engineering and technical personnel from domestic and foreign.There are still problems with the automatic parking system that it does not recognize irregular parking spaces and parking accuracy is not very precise.In order to solve the existing problems of automatic parking system,this thesis studies the key technology of automatic parking system based on multi-sensor information fusion to improve the intelligent degree of automatic parking system and the utilization rate of irregular parking space resources.This thesis mainly gives three aspects of parking identification technology,path tracking control technology,automatic parking system architecture design and function specification.In addition,the real vehicle experiment is used to test the design of the automatic parking system performance.First,the parking space identification technology was studied.Building a multi-sensor environment perception architecture,Ultrasonic radar detects the vertical depth of the parking spaces,the odometer measures the level of the parking spaces,and the camera is used to detect the wheels of the side docked vehicle.Through the multi-sensor information fusion,the establishment of parking spaces identification model,the use of fuzzy reasoning,learning from the skilled driver to manually select the parking spaces,so that parking spaces could automatically identify all parking spaces.Simulation analysis was conducted to verify the effectiveness of the design of the parking space identification technology.Secondly,the path tracking control technology was studied.For improving parking efficiency and parking accuracy,based on the global stability,a sliding mode variable structure control with hyperbolic tangent function is proposed.The tracking system of the vehicle kinematic model was transformed into a double closed-loop cascade subsystem with position controller and attitude controller,and a convergence law of exponential and power was designed to ensure the parking path achieves rapid convergence in a limited time,So that the parking tracing path could achieve rapid convergence in a limited time.Simulation analysis was conducted to verify the effectiveness of the designed path tracking techniqueThen,according to the demand of parking space recognition technology and path tracking control technology designed,build automatic parking system architecture.The environment-aware module,the decision-making module,the execution module and the communication module are designed in detail,analyzing the functional specifications of the four modules in each stage of automatic parkingFinally,the vehicle is modified according to the requirements of automatic parking system hardware architecture,for carrying out real vehicle experiments.The stability of the underlying electrical system and the perceptual performance of the environment-aware architecture to the environment are studied and tested.The calibration for the error of parking spaces caused by the process of searching for library is amended.The effectiveness of the parking space recognition technology based on multi-sensor information fusion and the tracking effect of the path tracking control technique based on sliding mode control are verified.Finally,the system suitability test is carried out.The experimental results show that the key technology of automatic parking system can identify complex irregular parking spaces and improve the parking accuracy,which can effectively improve the performance of existing automatic parking system.