论文/arXiv 2610.01612·27 天前

ReCo：用于腿式操作的带 Policy-Aware MPC 的响应一致性运动

ReCo通过响应整形与闭环响应模型结合MPC，实现行走中末端跟踪，并在仿真基准和真实实验验证。

AI 整理

• 本体 SDK 与控制
• 仿真与训练
• 具身模型

原文链接：ReCo: Response-Consistent Locomotion with Policy-Aware MPC for Legged Manipulation（arXiv 2610.01612 · 发布日期：2026-10-01）

ReCo: Response-Consistent Locomotion with Policy-Aware MPC for Legged Manipulation

ReCo通过响应整形与闭环响应模型结合MPC，实现行走中末端跟踪，并在仿真基准和真实实验验证。

核心要点

• 连续腿式操作需要底座行走时保持末端精确跟踪，且MPC只能补偿可预测的底座运动。
• ReCo通过响应整形训练策略，使其在随机动力学下对命令产生一致且可重复的响应。
• 识别出的闭环响应模型使MPC能够联合规划运动命令和机械臂运动，协调底座与手臂。
• 仿真基准上相较最佳基线位置和姿态RMSE分别降低28.7%和27.4%，真实实验展示板载连续腿式操作。

验证范围与局限：验证包括仿真基准和真实世界板载连续腿式操作实验。

英文摘要 (Abstract)

Continuous legged manipulation requires accurate end-effector tracking while the base keeps walking. Combining reinforcement learning (RL) with model predictive control (MPC) suits this task: the learned policy provides robust locomotion, while MPC coordinates the base and arm to compensate for tracking errors. However, MPC can compensate only for base motion that it can predict, and a learned policy's command response varies with gait phase, contact, and payload. We present ReCo, a framework that couples response-consistent locomotion with policy-aware MPC for legged manipulation. Response shaping trains the policy to respond to commands consistently and repeatably across randomized dynamics. An identified closed-loop response model then lets MPC jointly plan locomotion commands and arm motion. On the simulation benchmark, ReCo reduces position and orientation root-mean-square error (RMSE) by 28.7% and 27.4% relative to the best baseline for each metric. Real-world experiments demonstrate onboard continuous legged manipulation with coordinated base and arm motion.

作者：Kuankuan Sima, Yichao Gao, Chenxi Gu, Kefan Zhao, Lin Zhao

来源

本页事实对应的公开页面，以原文为准。

1. arXiv 2610.01612：ReCo: Response-Consistent Locomotion with Policy-Aware MPC for Legged Manipulation访问 2026-10-21