Reproduce “HWC-Loco: A Hierarchical Whole-Body Control Approach to Robust Humanoid Locomotion” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sixu Lin, Guanren Qiao, Yunxin Tai, Ang Li, Kui Jia, Guiliang Liu
HWC-Loco introduces a hierarchical whole-body control framework for robust humanoid locomotion, explicitly learning to recover from safety-critical situations while dynamically balancing task performance and safety. It achieves superior success rates on challenging terrains and under disturbances in both simulation and real-world deployment on Unitree H1 and G1 robots.
Key insights
- Formulating policy learning as a robust constrained Reinforcement Learning problem allows the system to explicitly learn recovery from safety-critical scenarios under extreme-case uncertainties, enhancing real-world robustness.
- A hierarchical control structure, composed of goal-tracking, safety recovery, and high-level planning policies, dynamically resolves the trade-off between task performance and stability, adapting behavior based on perceived risk.
- Integrating human motion imitation via Wasserstein distance and dynamic constraints like Zero Moment Point (ZMP) guidance ensures both natural, human-like movements and inherent stability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.