arXiv 2511.211692025-11-26cs.RO

具备力交互能力的人形机器人多策略强化学习

Kinematics-Aware Multi-Policy Reinforcement Learning for Force-Capable Humanoid Loco-Manipulation

the Robot and Artificial Intellifence Lab, College of Electronic and Information Engineering, Tongji University, Shanghai 201804, China（机器人与人工智能实验室，电子与信息工程学院，同济大学，上海）；State Key Laboratory of Autonomous Intelligent Unmanned System(Tongji University), Shanghai, China（自主智能无人系统国家重点实验室（同济大学），上海）

Kaiyan Xiao, Zihan Xu, Cheng Zhe, Chengju Liu, Qijun Chen

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出了一种多策略强化学习框架，通过解耦的三阶段训练流程提升人形机器人在高负载工业场景中的运动-操作能力。

摘要：

人形机器人因其类人形态，在工业应用中具有巨大潜力。然而，现有运动-操作方法主要专注于灵活操作，无法满足高负载工业场景中对灵活性和主动力交互的综合需求。为此，我们提出了一种基于强化学习的框架，包含一个解耦的三阶段训练流程，包括上半身策略、下半身策略和delta命令策略。为加速上半身训练，设计了一种启发式奖励函数。通过隐式嵌入前向动力学先验，使策略能够更快收敛并实现更优性能。对于下半身，开发了一种基于力的课程学习策略，使机器人能够主动施加并调节与环境的交互力。

英文摘要：

Humanoid robots, with their human-like morphology, hold great potential for industrial applications. However, existing loco-manipulation methods primarily focus on dexterous manipulation, falling short of the combined requirements for dexterity and proactive force interaction in high-load industrial scenarios. To bridge this gap, we propose a reinforcement learning-based framework with a decoupled three-stage training pipeline, consisting of an upper-body policy, a lower-body policy, and a delta-command policy. To accelerate upper-body training, a heuristic reward function is designed. By implicitly embedding forward kinematics priors, it enables the policy to converge faster and achieve superior performance. For the lower body, a force-based curriculum learning strategy is developed, enabling the robot to actively exert and regulate interaction forces with the environment.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。