Automatic Parallel Parking Method Of Distributed Drive Electric Vehicle Based On Vision

With the substantial increase in the number of cars,cars not only bring convenience to people,but also make urban traffic congestion and parking more difficult.Automatic parking technology is one of the best means to solve this problem,and it is also one of the important functions of intelligent connected cars.In the past two decades,the technology of automatic parking has developed rapidly.Most of the current environmental sensing sensors are ranging sensors such as ultrasonic radars,which cannot identify environmental information such as planes.Most of the control algorithms are based on the experience of driver control.Aiming at the issue of appeal,this paper studies the vision sensor-based automatic parking sensing system,and simultaneously performs global path planning and tracking for automatic parallel parking.Distributed driving electric vehicle is the ultimate development direction of pure electric vehicles.It has obvious control advantages in terms of handling stability and active safety.It has easy controllability,high efficiency,and simple structure.It separates the driving torque of distributed driving electric vehicles independently.Controllable characteristics are applied in the research of automatic parking methods.Differential steering assisted active steering reduces the lateral control error of global path tracking in automatic parking technology.This article relies on the National Natural Science Foundation of China(51675257)and the Innovative Talents Project of Higher Education Institutions in Liaoning Province(LR2016054)to conduct research on the vision-based distributed parallel automatic parking of electric vehicles.The thesis mainly includes the following four parts:(1)Design of panoramic surround view perception system.Four fisheye cameras are arranged around the vehicle body to realize the panoramic perception of the vehicle body.The four fisheye phases are calibrated using the Zhang Zhengyou calibration method through simulation scenes,and then converted into a bird's-eye view through inverse perspective transformation.(2)Parking space detection based on customized Radon transform of the scene.Firstly,the panoramic surround view perception image is pre-processed,and edge features are extracted by Canny operator.Then,the scene-based customized Radon transform is used to detect the parking space edge according to the characteristics of the automatic parallel parking scene.(3)Automatic parallel parking path planning.The parking path is divided into two phases,which are composed of two arcs.A tentacle algorithm is introduced to pre-set the parking route library,and the tentacles that meet the requirements are filtered through rule constraints.It is combined with the second section of the parking path design based on the arc method.The parking path reduces the requirements for the initial parking posture.Simulation experiments verify that the planned path can realize the automatic parallel parking function.(4)Tracking compensation for differential steering of distributed drive electric vehicles.When the vehicle cannot be parked completely according to the turning radius of the planned path due to external factors,the yaw moment is generated by the distribution of the driving force of the left and right wheels to achieve differential steering to compensate the tracking error.The research results show that the vision-based distributed parallel drive electric vehicle automatic parallel parking method can make the distributed drive electric vehicle identify the parking space and plan the parking path of the target parking space.At the same time,when there is a deviation between the actual turning radius and the target turning radius,it can be compensated by differential steering to reduce the steering deviation.This research has certain reference value for the development of future ADAS control methods.