<ul data-eligibleForWebStory="true"><li>Humanoid robots are valuable for daily tasks due to their flexibility and human-like features.</li><li>Previous methods for whole-body control and loco-manipulation in humanoids require task-specific tuning.</li><li>SkillBlender is a new hierarchical reinforcement learning framework for versatile humanoid loco-manipulation.</li><li>SkillBlender pretrains task-agnostic primitive skills and blends them dynamically for complex tasks.</li><li>SkillBench is introduced as a benchmark with diverse embodiments, skills, and tasks for evaluation.</li><li>Extensive simulated experiments show SkillBlender outperforms baselines in loco-manipulation tasks.</li><li>SkillBlender also prevents reward hacking and produces accurate and feasible movements.</li><li>The project code and benchmark will be open-sourced to support future research.</li></ul>

SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending

Discover more