Hierarchical Residual Policy Learning for Real-World Mobile Manipulation with Sparse Human Guidance
arXiv cs.AI · 15 Oct 2026, 03:59 AEST
HiRe-MoMa is a new framework designed for efficient fine-tuning of mobile manipulation systems in real-world settings. It utilizes a hierarchical architecture to enhance task performance with minimal human guidance.
Key facts
- 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.
- At its core, HiRe-MoMa adopts a Hierarchical Residual Architecture that fine-tunes a frozen base policy.
- 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.
- Extensive real-world experiments on the TIAGo++ platform demonstrate that HiRe-MoMa consistently enhances task performance with only around 30 minutes of online interaction.
Read aloud
HiRe-MoMa is a new framework designed for efficient fine-tuning of mobile manipulation systems in real-world settings. It utilizes a hierarchical architecture to enhance task performance with minimal human guidance. 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. At its core, HiRe-MoMa adopts a Hierarchical Residual Architecture that fine-tunes a frozen base policy. To improve sample efficiency in data-expensive real-world settings, we propose Expert-Guided Value Alignment, which leverages sparse human interventions with under twenty percent training coverage to reshape the critic's value landscape. Extensive real-world experiments on the TIAGo++ platform demonstrate that HiRe-MoMa consistently enhances task performance with only around thirty minutes of online interaction.
Sources in this cluster
- Hierarchical Residual Policy Learning for Real-World Mobile Manipulation with Sparse Human Guidance
arXiv cs.AI · 15 Oct 2026, 03:59 AEST