基于环境风险的自动驾驶局部路径规划算法

Local Path Planning Algorithm for Autonomous Driving Based on Environmental Risk

针对结构化道路局部路径规划问题,提出基于实时环境风险场的自动驾驶局部路径动态规划框架,首次将局部路径规划问题细化为车道决策和路径规划两部分。针对车道决策部分,提出基于行车风险场及安全距离的车道决策算法,能够在保证驾驶速度的同时,确保自车始终处于低风险驾驶环境,以提高其安全性。在换道路径规划部分,提出基于换道时间均匀采样的候选路径生成算法,以及综合考虑换道即时性、速度平滑性、路径平顺性以及舒适性的代价函数,实现最优路径规划。在单车道路径规划部分,提出综合考虑安全性、平顺性以及连续性的代价函数,实现路径以及速度的合理、安全动态规划。实验验证表明,所提出的局部路径动态规划框架在设定的结构化道路局部路径规划任务中,能够规划出低风险、高效率、安全合理且平顺的行驶路径并给出安全规划速度,证明了所提出的局部路径动态规划算法的有效性。

In this study,we propose a dynamic local path planning framework for autonomous driving based on real-time environmental risk field so as to resolve the problem of local path planning of structured roads.To the best of our knowledge,the local path planning problem is categorized into lane decision and path planning for the first time.With respect to lane decision-making,a lane decision-making algorithm based on driving risk field and safety distance is proposed to maintain the driving speed of the self-vehicle and ensure that the vehicle is always in a low-risk driving environment to improve its safety.With respect to lane change path planning,a candidate path generation algorithm based on the uniform sampling of lane change time is proposed.In addition,we realize optimal path planning by proposing a cost function that comprehensively considers the instantaneity of lane change,speed smoothness,path smoothness,and comfort.Regarding single-vehicle path planning,we propose a cost function that considers safety,smoothness,and continuity to realize reasonable and safe dynamic planning of path and speed.Our experimental results show that the proposed local path dynamic planning framework can plan a low-risk,efficient,safe,reasonable,and smooth driving path and give a safe planning speed in the set local path planning task of structured roads.Thus,in other words,our results prove the effectiveness of the proposed local path dynamic planning algorithm.