为提高驾乘人员对自动驾驶车辆换道轨迹的接受度,解决轨迹规划中路径与速度解耦的问题,以高速公路自然驾驶数据集为依托,提出一种基于动态运动基元(Dynamic movement primitives,DMPs)的车辆高速公路换道轨迹规划方法.首先,利用极大似然估计对车辆高速公路历史换道轨迹信息进行分析与处理,得到能够表征自然驾驶员换道特性的换道示教轨迹;其次,采用DMPs算法对换道示教轨迹进行学习,完成对示教轨迹的复现并构建运动基元库;之后,基于运动基元库及换道需求,可泛化出多条适应高速公路驾驶场景的换道轨迹;最后,基于Prescan/Matlab/Carsim联合仿真和实车试验对规划轨迹的实时性和可跟踪性进行验证.结果表明,所提出的换道轨迹规划算法具有计算效率高、泛化能力强、驾乘人员接受度好的优点,并实现了换道路径与速度的协同规划.
Expressway Lane Changing Trajectory Planning Method of Autonomous Vehicle Based on Dynamic Movement Primitives
To enhance drivers'acceptance of autonomous vehicles while also addressing the decoupling path and velocity in trajectory planning,an expressway lane-changing trajectory planning method based on dynamic movement primitives(DMPs)is proposed utilizing natural expressway driving data.First,the expressway's historical lane change trajectory is analyzed and processed using the maximum likelihood estimation to obtain the teaching lane change trajectory,which can characterize the natural driver's lane change characteristics.Second,the DMPs algorithm is adopted to learn the teaching lane change trajectory,complete the trajectory replication,and construct a library of movement primitives.Then,based on the library of movement primitives and lane change requirements,multiple lane change trajectories can be generalized that can suit expressway driving scenarios.Finally,the planned trajectory's real-time and traceability is verified based on the joint Prescan/Matlab/Carsim simulation and vehicle experiment.The results demonstrate that the proposed DMPs algorithm offers efficient computation,excellent generalization capability,and high acceptance by drivers.Moreover,it enables cooperative planning of lane change path and velocity.
autonomous vehiclelane changing trajectory planningdynamic movement primitivesteaching trajectory
NETL
NSTL
万方数据
