Extreme-RGMT: Continual Learning of Highly Dynamic Skills for Robust Generalist Humanoid Control

2026-07-22

AI Summary

This work addresses the challenge that existing humanoid robot controllers struggle to simultaneously achieve general locomotion capabilities and learn high-dynamic skills, often suffering from catastrophic forgetting when acquiring new motions. To overcome this, the authors propose a two-stage continual learning framework: first, a general base policy is trained using diverse motion data; subsequently, an asymmetric skill acquisition and capability consolidation mechanism is introduced, augmented with difficulty-aware sampling and advantage-prioritized trajectory resampling. This approach effectively mitigates the scarcity of high-dynamic motion samples and their high failure rates during training. The method significantly improves success rates in executing high-dynamic actions, achieves human-expert-level performance across a wide range of unseen tasks, and sets a new state of the art in general full-body motion tracking.

Abstract

Humans can progressively acquire highly dynamic motor skills while preserving reliable everyday motor abilities. In contrast, existing humanoid controllers face a trade-off between generalist and specialist capabilities: generalist motion tracking policies struggle to reliably execute rare highly dynamic motions, whereas specialist training can degrade previously acquired behaviors. We introduce Extreme-RGMT, a two-stage continual learning framework for robust generalist humanoid control. The method first learns a generalist motion-tracking base policy from diverse multi-source motion data, then employs an asymmetric skill acquisition and capability consolidation mechanism to constrain policy drift on mastered motions while emphasizing difficult dynamic segments. To address the scarcity of highly dynamic motions, their high failure rates, and the resulting shortage of informative samples, Extreme-RGMT combines difficulty-aware sampling with advantage-prioritized trajectory resampling to emphasize critical segments. Experiments show that Extreme-RGMT achieves state-of-the-art generalist whole-body motion-tracking performance, including substantially improved completion of challenging highly dynamic motions. The resulting controller directly executes diverse unseen highly dynamic motions under fixed references and online inertial motion-capture inputs, advancing generalist whole-body motion-tracking controllers toward highly dynamic motor capabilities at the human-expert level.

Problem

• continual learning
• humanoid control
• highly dynamic skills
• generalist policy
• motion tracking

Innovation

• continual learning
• humanoid control
• motion tracking
• dynamic skills
• policy consolidation