The Path Palnning Decision Making For The Low Speed Automated Driving

In recent years,with the development of the automated driving vehicle,people have realized that the automated driving vehicle can play a great role in transportation,logistics transportation.It can accomplish tasks more efficiently and quickly.The path planning is the most basic and important role in the research of automated driving vehicle.The main research contents of this paper are global path planning and local path planning.In the closed field,local path planning based on global path planning can achieve the automated driving function in closed field.The specific research contents are in the following aspects:The closed site map is processed,and the transformation between geodetic coordinate and plane rectangular coordinates is completed by positive and negative operations of the Gauss transformation,and map information is transformed into network model information.The Dijkstra's algorithm is used to the global path planning.When give starting point and the terminating point,the optimal path of the starting point to the end point is planned.And a series of trajectory points are generated in the optimal path so that the vehicle can be driven automatically according to the trajectory point.The local path planning is carried out on the basis of the global path planning.When the vehicle runs by the global path planning,the local path planning will be carried out when the obstacle is encountered in front of the vehicle.The multi-objective reinforcement learning method is applied to local path planning.And comparing with the local path planning based on single point preview model,the advantage of this research method is obtained.Finally,aiming at the large scale of the multi object reinforcement learning method and the slow operation time,the method of fitting multi-objective reinforcement learning is proposed.The Cubic Bezier curve is used to fit the nodes of the multi-objective enhanced learning method.There is great improvement in the path generation time of local path planning.Simulate by using global path planning based on Dijkstra's algorithm and local path planning based on reinforcement learning by Matlab and do road experiment.And finally verify the practicability of the method in the actual closed road.