| With the development of autonomous driving technology,automatic parking（SAE L2）and its evolution direction autonomous parking（SAE L4）have become the solution to this problem.Autonomous parking system（APS）has two technical routes,one is off-field end scheme,the other is off-field end scheme.The end intelligent scheme has high requirements on vehicle intelligence,and the current technology is not mature.Its L2 level fully autonomous parking system is only applicable to high-end cars and electric cars,but not really applicable to ordinary cars.In view of the above situation,the selection of field end intelligence has great advantages.It can fill the shortcomings of common vehicles’ lack of intelligence through the transformation of infrastructure and communication technology,and at the same time,it is relatively easy to achieve technically.Therefore,this paper proposes the automatic parking service based on vehicle-road cloud collaboration,and studies the relevant technologies of automatic parking with middle and low-end vehicles as service objects.The main contents are as follows:1.Determine the cloud framework structure of the automatic parking system according to the modification of experimental vehicles and roadside equipment in the parking lot,and determine the realization of the two functions of perception fusion and positioning and global path planning in the cloud.2.Collect the laser point cloud map of the parking lot using experimental vehicles,and then build the map through laser SLAM to achieve high-precision map collection and production.Complete the functions of the cloud perception fusion and positioning module,including the fusion of multi-source perception information transmitted to the cloud by the vehicle end,road and side devices and the conversion from geographic coordinates to plane coordinates.3.In view of the situation that the system selects parking Spaces for users independently,considering the factors of distance from vehicle position to parking space,parking space to exit and whether there is parking space on both sides of the parking space,a parking space selection algorithm with equal emphasis on convenience and comfort is designed.4.Study the methods of path planning,analyze the advantages and disadvantages of different methods,and determine the method to choose in the experiment.A two-layer cloud path planning method is proposed for large-scale maps in the park.A* algorithm based on topology map is used to select paths in the park at the upper level,and A* algorithm based on grid map is used to plan paths between nodes in the road network at the lower level.5.Of the lower hybrid A * algorithm used in the planning,to research the composition of the cost function and heuristic function part of them was improved,join in the heuristic function about parent node constraint,make it A better convergence towards the target and the generated path can be more smooth,can better application in automatic parking application,Finally,a real vehicle test was carried out. |
