Reproduce “HiLMa-Res: A General Hierarchical Framework via Residual RL for Combining Quadrupedal Locomotion and Manipulation” 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
Xiaoyu Huang, Qiayuan Liao, Yiming Ni, Zhongyu Li, Laura Smith, Sergey Levine, Xue Bin Peng, Koushil Sreenath
HiLMa-Res coordinates quadruped walking and leg-based manipulation using a reusable controller that tracks adjustable toe paths, while task-specific planners choose path adjustments and base commands. In real-world NavLoad trials, its real-world-trained policy achieved 100% success and 47.6-second average time, versus 80% and 57.75 seconds for its simulation-trained policy. These results concern NavLoad, not all tasks or gaits.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.