外观
论文arXiv 2609.189301 个月前
使用双足移动机械臂学习整体式全身移动操作
原文链接：Learning Holistic Whole-Body Loco-Manipulation with a Bipedal Mobile Manipulator（arXiv 2609.18930 · 发布日期：2026-09-16）
Learning Holistic Whole-Body Loco-Manipulation with a Bipedal Mobile Manipulator
该工作用强化学习训练统一全身控制器，将6-DoF末端目标映射为臂腿协同动作，并在真机多种命令来源下验证。
核心要点
- 双足移动操作需协调移动与操作，使机器人触及手臂常规工作空间之外的物体。
- 控制器仅给定末端目标，自主协调伸手、姿态适应与迈步，无需显式底盘速度或落脚命令。
- 训练采用奖励门控平衡末端跟踪、移动与平衡，并用时序上下文估计器提取动态相关信息。
- 真机实验显示同一控制器可接入VR遥操作、学习的Diffusion Policy和脚本轨迹命令。
验证范围与局限：验证范围为真机上的伸手、姿态适应和迈步，以及VR遥操作、学习的Diffusion Policy和脚本轨迹命令。
英文摘要 (Abstract)
Bipedal loco-manipulation enables robots to interact with objects beyond the nominal workspace of their arms by coordinating locomotion and manipulation. Realizing this capability requires a low-level whole-body controller that translates task-level manipulation goals into coordinated arm and leg motions while maintaining balance. We present a unified whole-body controller trained with reinforcement learning that directly maps 6-DoF end-effector targets to coordinated actions for the bipedal base and robotic arm. Given only an end-effector target, the learned controller autonomously coordinates reaching, postural adaptation, and stepping without explicit base-velocity or footstep commands. A reward-gating strategy regulates the trade-offs among end-effector tracking, locomotion, and balance during training, while a temporal context estimator combines windowed Transformer encoding, recurrent GRU memory, and auxiliary dynamics prediction to extract dynamics-relevant information from observation history. Real-robot experiments demonstrate that the same controller supports reaching, postural adaptation, and stepping under commands from VR teleoperation, a learned diffusion policy, and scripted trajectories, providing a common end-effector interface for diverse manipulation tasks.
作者：Zhongyu Chen, Yuxuan Nai, Qian Chen, Yidong Zhu, Chen Jing, Qihan Wang, Xudong Li, Zhizhan Li, Leixin Chang, Liangjing Yang
来源
本页事实对应的公开页面，以原文为准。