This paper presents a hierarchical replanning framework that rapidly modulates the ongoing trajectory to help 7-DoF redundant manipulators avoid dynamic obstacles in physical human-robot interaction. In this method, the safety requirements of the end-effector and links are met without adding additional control burden. Firstly, an improved rapidly-exploring random trees planner is utilized to initialize the path in the joint space. To prevent excessive trimming, a heuristic method based on geometric path descriptors is proposed for finding and removing redundant nodes. Secondly, a hybrid scheme combining local path rewiring and redundancy-based node self-motion is used to identify and replan the part of the pre-planned path affected by dynamic obstacles. Lastly, an adaptive online trajectory generator is designed to adjust the trajectory in real time and generate smooth motion primitives for actuators. Simulations and real-world experiments validate the effectiveness and robustness of the proposed framework for whole-arm obstacle avoidance in dynamic scenarios. Compared to other methods, our approach has a 35% higher success rate and a 47.2% reduction in replanning time.

本文提出了一种分层重新规划框架，该框架可快速调节正在进行的轨迹，以帮助 7-DoF 冗余机械手避免物理人机交互中的动态障碍。在这种方法中，在不增加额外控制负担的情况下满足末端执行器和链接的安全要求。首先，利用改进的快速探索随机树规划器来初始化联合空间中的路径。为了防止过度修剪，提出了一种基于几何路径描述符的启发式方法来查找和去除冗余节点。其次，结合局部路径重新布线和基于冗余的节点自运动的混合方案用于识别和重新规划受动态障碍影响的预先规划的路径部分。最后，自适应在线轨迹生成器旨在实时调整轨迹并为执行器生成平滑的运动原语。模拟和现实世界的实验验证了所提出的框架在动态场景中全臂避障的有效性和鲁棒性。与其他方法相比，我们的方法成功率提高了 35%，重新规划时间减少了 47.2%。