高速公路无人驾驶的分层抽样多动态窗口轨迹规划算法  

Stratified Sampling Based Multi-dynamic Window Trajectory Planner for Autonomous Driving on Highway

作　　者：张琳 薛建儒[1] 马超 李庚欣 李勇强 ZHANG Lin;XUE Jian-Ru;MA Chao;LI Geng-Xin;LI Yong-Qiang(Institute of Artificial Intelligence and Robotics,Xi'an Jiaotong University,Xi'an 710049)

机构地区：[1]西安交通大学人工智能与机器人研究所,西安710049

出　　处：《自动化学报》2024年第7期1315-1332,共18页Acta Automatica Sinica

基　　金：国家自然科学基金(62036008,61773311)资助。

摘　　要：高速公路无人驾驶轨迹规划面临着实时性强、安全性高的挑战.为此,提出一种分层抽样多动态窗口的轨迹规划算法(Stratified sampling based multi-dynamic window trajectory planner,SMWTP).首先,用多动态窗口表征可行轨迹的搜索空间,并基于贝叶斯网络构建轨迹概率分布模型.其次,采用先速度后路径的分层抽样策略生成符合动态场景约束的候选轨迹集合.最后,利用引入障碍车辆速度估计不确定性的责任敏感安全模型(Responsibility sensitive safety,RSS)从中选择最优轨迹.大量仿真实验和实际交通场景测试验证了算法的有效性,对比实验结果表明,所提算法性能显著优于人工势场最优轨迹规划算法和多动态窗口模拟退火轨迹规划算法.Autonomous driving trajectory planning on highways faces challenges of strong real-time performance and safety.This paper proposes a stratified sampling based multi-dynamic window trajectory planner(SMWTP)for unmanned vehicles on highway.Firstly,the search space of feasible trajectories is constructed with multi-dynamic windows.Then,the Bayesian network is used to derive the probability distribution model of trajectories.Secondly,the stratified sampling strategy where speed is sampled before path makes generated candidate trajectories meet the constraints in dynamic scenes.Finally,the uncertainty of traffic participant vehicles'speed estimation is embedded into responsibility sensitive safety(RSS)model to select the optimal trajectory.A large number of simulation experiments and real traffic scenario tests have verified the effectiveness of the algorithm.The comparative experimental results show that the performance of the proposed algorithm is significantly better than the optimal trajectory planning algorithm based on artificial potential fields and multi-dynamic window simulated annealing-optimized trajectory planning algorithm.

关 键 词：无人驾驶 轨迹规划 运动规划 贝叶斯网络 

分 类 号：U463.6[机械工程—车辆工程] TP18[交通运输工程—载运工具运用工程]