arXiv：cs.LG· Zhefei Gong, Hefeifei Jiang, Zehao Wang, Jiachen Li·· 2 小时前
HiRe-MoMa：面向真实世界移动操作的分层残差策略学习框架
Hierarchical Residual Policy Learning for Real-World Mobile Manipulation with Sparse Human Guidance
AI 导读
研究提出 HiRe-MoMa，一个针对复杂移动操作（MoMa）系统的分层残差框架，通过冻结基础策略并叠加策略无关残差，结合移动底座与机械臂头的分层解耦来缓解梯度干扰。其专家引导价值对齐方法利用覆盖率低于 20% 的稀疏人工干预重塑 critic 价值分布，在 TIAGo++ 平台上仅需约 30 分钟在线交互即可稳定提升任务表现，且不受基础策略架构限制。
正文
Abstract:Mobile Manipulation (MoMa) is a cornerstone of autonomous robotic systems, yet real-world fine-tuning remains a formidable challenge because of high-dimensional action spaces and intricate coupling between heterogeneous modules. In this paper, we propose HiRe-MoMa (Hierarchical Residual Mobile Manipulation), a systematic framework for stable and sample-efficient real-world fine-tuning of complex MoMa systems. At its core, HiRe-MoMa adopts a Hierarchical Residual Architecture that fine-tunes a frozen base policy. A Policy-Agnostic Residual learns additive corrections atop the base policy, while Hierarchical Decoupling splits the residual into mobile-base and manipulator-arm heads with unidirectional conditioning, mitigating gradient interference and aligning corrections with the platform's physical hierarchy. To improve sample efficiency in data-expensive real-world settings, we propose Expert-Guided Value Alignment, which leverages sparse human interventions with under 20% training coverage to reshape the critic's value landscape, effectively guiding autonomous exploration through expert priors. Extensive real-world experiments on the TIAGo++ platform demonstrate that HiRe-MoMa consistently enhances task performance with only around 30 minutes of online interaction, regardless of the base policy architecture. To our knowledge, this is among the first frameworks to enable efficient on-robot fine-tuning for mobile manipulation in the real world, providing a practical paradigm for deploying complex robotic systems in dynamic, human-centric environments. Further details can be found on our project website: this https URL
| Comments: | Conference on Robot Learning (CoRL 2026); Project Website: this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.20607 [cs.RO] |
| (or arXiv:2610.20607v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20607 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhefei Gong [view email]
[v1]
Wed, 14 Oct 2026 17:59:52 UTC (5,982 KB)
来源：arXiv：cs.LG · arxiv.org