基于改进人工势场的自动驾驶汽车弯道超车动态路径规划

动态路径规划是自动驾驶汽车避障控制的关键技术。针对自动驾驶汽车弯道超车工况，建立基于改进人工势场(Artificial Potential Field, APF)的动态路径规划方法。为使基于APF的动态路径规划方法能运用于包含弯曲道路的复杂交通环境，将已在直道环境验证过的道路APF函数通过极坐标系与笛卡尔坐标系的相互转换，建立考虑道路曲率的弯曲道路APF函数。针对根据车辆质心位置判断车辆碰撞风险方法存在的缺陷，提出考虑车辆体积的碰撞风险预判方法，建立综合考虑车辆位置、速度和体积的障碍车辆APF函数。基于弯曲道路APF和改进障碍车辆APF，建立道路环境综合APF，引导车辆实现弯道超车。为避免目标函数中子目标相互干涉，提高弯道超车安全性，提出根据本车与障碍车辆相对位置关系自适应调整权重矩阵的方法。基于Carsim/Simulink联合仿真平台，分别在静态障碍车辆和动态障碍车辆2种工况下，验证自动驾驶汽车弯道超车动态路径规划的有效性。研究结果表明：所建立的弯曲道路APF能引导车辆转弯行驶，避免冲出车道；目标函数权重自适应调整方法能根据超车过程动态调整子目标的权重，规划出符合道路交通安全法规的路径，避免车辆超车时提前折返原车道，提高了超车安全性；考虑车辆体积的障碍车辆APF提高了车辆碰撞风险的预判精度，有效避免碰撞事故发生。

Dynamic Path Planning for Self-driving Cars Overtaking on Curves Based on Improved Artificial Potential Field

Dynamic path planning is the key technology in obstacle avoidance of self-driving cars. A dynamic path planning method for the self-driving cars under overtaking on curves based on improved artificial potential field (APF) was proposed. To apply the dynamic path planning method based on APF to the complex traffic environment including curved road, the APF function which had been verified in straight road environment was transformed by the polar coordinate system and Cartesian coordinate system, and the curved road APF function considering road curvature was established. Aiming at the defects of the method of judging vehicle collision risk according to the position of vehicle centroid, a collision risk prediction method was proposed, and then the APF function of obstacle vehicle was established where the vehicle position, velocity and volume were comprehensively considered. Thus, a comprehensive APF of road environment was established based on the APF of the curved road and the obstacle vehicles, guiding the vehicle to overtake on the curve. To avoid the interference between sub objectives in the objective function and improve the overtaking safety on curves, a method of adaptively adjusting the weight matrix according to the relative position between the vehicle and the obstacle vehicle was proposed. Based on the Carsim/Simulink co-simulation platform, the validity verification of the dynamic path planning for the self-driving cars overtaking on curves was carried out under static and dynamic obstacle vehicle conditions, respectively. The results show that the APF of curved road can guide vehicles to turn and avoid rushing out of the lane; the adaptive adjustment method of objective function weight can dynamically adjust the weight of sub objectives according to the overtaking process, plan the path in line with road traffic safety regulations, avoid turning back to the original lane before overtaking the ahead car, and improve the overtaking safety; the APF of obstacle vehicle considering vehicle volume improves the prediction accuracy of vehicle collision risk and effectively avoids collision accidents.