Multi-humanoid coordination improves with skill blending framework
MASkillBlender: Decentralized Whole-Body Coordination for Multi-Humanoid Loco-Manipulation via Skill Blending
RoboticsMachine LearningMultiagent Systems
Summary
Controlling several humanoid robots to work together on tasks is very hard because it involves managing many complex body movements and decisions without a central controller. The authors introduce MASkillBlender, a new approach that helps multiple humanoid robots coordinate by combining pre-learned skills, using only high-level goals without needing detailed instructions for each task. They also add a method to make learning faster for groups of similar robots, preserving important learning information. Tests in simulations show their method helps robots work well together across various tasks and robot types.
What this means in practice
- •For robotics engineers: Create decentralized control systems enabling multiple humanoid robots to coordinate complex movement and manipulation tasks efficiently.
- •For robotics simulation developers: Build simulation environments to train and test multi-humanoid coordination strategies that generalize across tasks and robot models.
Authors
Yifan Hu, Luhang Hong, Mingkang Long, Danning Wang, Chengfeng Jia, Rong Su, Junjie Fu, Guanghui Wen
Abstract
Coordinated multi-humanoid loco-manipulation is promising yet challenging due to high-dimensional whole-body control, decentralized decision making, and scalability. While recent reinforcement learning methods have improved single-humanoid whole-body control, extending them to the multi-humanoid setting remains nontrivial and often requires substantial reward engineering or task-specific design. We propose MASkillBlender, a general multi-agent reinforcement learning framework to achieve decentralized multi-humanoid whole-body coordination. By learning a shared decentralized high-level policy over reusable pre-trained single-humanoid skills, MASkillBlender enables coordinated behaviors using only task-level rewards, without requiring task-specific motion references. To improve learning efficiency, we further introduce a permutation-based data augmentation strategy for homogeneous multi-humanoid systems, and theoretically show that the permuted samples preserve the policy-gradient direction of the original samples under the homogeneous Markov game formulation. We evaluate MASkillBlender on multiple multi-humanoid coordination tasks across two humanoid embodiments. Simulation results demonstrate that the proposed framework consistently achieves strong task performance and enables coordinated behaviors across different tasks and humanoid embodiments.