Research On Automatic Parking Path Planning And Control Strategy Based On Panoramic Technology

With the increasing quality of living standards,my country’s total amount of vehicles is also increasing year by year,and the problem of parking difficulties in cities is not uncommon.As one of the representative products of the new era of automobile intelligent driving technology,the automatic parking system can help drivers greatly reduce the parking pressure,solve problems such as some urban road congestion,safe traffic accidents caused by parking,etc.,so that people can truly experience to the benefits brought by the advancement of vehicle intelligent driving technology.The automatic parking system integrates perception,planning,positioning,and control execution,which can not only improve the safety and reliability of the parking process,but also promote the progress of vehicle intelligent driving technology,which has wide application prospects and research value.In view of the popular application of panoramic technology in vehicles,this paper researchs the vital technologies of automatic parking system,such as parking space detection,path planning,motion control,etc.,on its basis.The specific research contents are as follows:First of all,this thesis introduces the parking space detection technology on the basis of around view monitor system,using fisheye camera as the sensing element,according to the problems of poor detection accuracy and weak generalization ability of traditional parking space detection methods,a deep convolution suitable for parking empty parking space detection is designed.The neural network is trained and tested using the classic parking space dataset.The dataset consists of more than 10,000 fisheye camera images of different colors,including three types of parking slots:parallel parking slots,vertical parking slots,and oblique parking slots.Through the detection of "L" or "T" type parking space markers,geometric relationship analysis and image classification,the entire empty parking space is determined.After training and experiments on the dataset,the outcomes indicate that the designed model has good performance such as high detection accuracy and strong generalization ability.Then,the kinematic geometric model of the vehicle at slow velocity is simplified by the relationship based on the Ackerman steering transformation.The single-step parking process is analyzed,and the minimum necessary parking slot parameters under the parallel parking space are calculated.And use the reverse reasoning method to preliminarily plan the parking path,and calculate the parking beginning position under the three types of parking spaces.Using the polynomial method to plan the subsequent parking path,and analyze the constraints in the parking path planning process,including vehicle kinematic constraints,environmental constraints,etc.,so that a parking path that satisfies the continuous curvature and meets the requirements of security can be obtained.Using Matlab to simulate and analyze the polynomial parking path planning for three types of parking spaces,and the path planning method is verified.Secondly,the sensor fusion technology is used to analyze the positioning problem within the vehicle range.The mileage information of the vehicle is obtained by analyzing the wheel track deduction algorithm.the relatively accurate pose information is obtained by calibrating according to the working principle of the inertial navigation system IMU.The extended Kalman filter algorithm is studied on the robot operating system,and the fusion positioning of the vehicle can be realized by the combined the mileage information with the pose data obtained by the IMU sensor.The feasibility of the positioning method is demonstrated by the experiment of the car built by the wirecontrolled chassis.Finally,by constructing the joint simulation experiment platform of Simulink and Carsim,the performance analysis of the maked path tracking controller based on model predictive control and pure tracking algorithm is carried out respectively.It can be seen from the simulation outcomes that the two control methods have completed the tracking control of the parking planning route,and the performance of the path tracking controller base on MPC is better than the path tracking controller based on pure tracking control.