基于改进人工势场法的山区公路自动驾驶车辆路径规划研究

Study on Path Planning of Mountain Roads Auto-Driving Vehicle Based on Improved Artificial Potential Field Method

针对目前的人工势场法应用于山区公路环境中的局限性,如主车与目标点的空间距离小于实际行驶距离而导致局部最小值的出现、障碍车辆斥力作用域不能随道路曲率的改变而变化导致干扰相邻车道中主车正常的行驶轨迹、双向车道中主车借道避障行为发生后不能及时返回原车道而保持逆行等问题,基于人工势场法原理,建立了改进的自动驾驶车辆人工势场模型路径规划,提出了分段引力势场模型和障碍物随动斥力势场模型,同时引入车道边界梯度斥力势场模型,在结构化山区道路环境中进行避障路径规划。Matlab软件仿真结果表明:在双向行驶车道工况下,主车借道避开障碍物,驶离障碍物斥力作用域后将及时回到原车道,避免长时间逆向行驶;在平曲线车道工况下,主车将保持当前车道正常行驶,不会受到相邻车道正常行驶的车辆干扰主车的行驶轨迹;模拟结果中并未出现局部最小值,改进的人工势场模型能够规划出符合汽车实际行驶过程中的安全避障路径。

The current artificial potential field method has limitations when applied to the mountain road environment.For example,the spatial distance between the main vehicle and the target point is less than the actual driving distance,which leads to the appearance of a local minimum.The scope of the obstacle vehicle repulsion force cannot change with the change of the road curvature,leading to interference with the normal driving trajectory of the main vehicle in the adjacent lane,and the main vehicle in the two-way lane cannot return to the original lane in time after the obstacle avoidance behavior occurs and maintain the retrograde.Based on the principle of artificial potential field method,an improved artificial potential field model path planning for autonomous vehicles is established,and a segmented gravitational potential field model and an obstacle follow-up repulsive force potential field model are proposed.At the same time,the gradient repulsion potential field model of the lane boundary is introduced to plan the obstacle avoidance path in the structured mountain road environment.The simulation results of Matlab software show that in the two-way driving lane conditions,the main vehicle will avoid obstacles by borrowing the lane,and will return to the original lane in time after leaving the obstacle repulsion range to avoid long-term reverse driving;in the flat curve lane conditions,the main vehicle will maintain the normal driving in the current lane,and will not be interfered with by the normal driving vehicles in the adjacent lane;There is no local minimum in the simulation results,and the improved artificial potential field model can plan a safe obstacle avoidance path in line with the actual driving process of the car.