H^2MARL whole-body control framework for humanoid robots released
What happened
Xuemin Hu and other authors submitted a preprint paper proposing H^2MARL, a hierarchical heterogeneous multi-agent reinforcement learning framework for whole-body control of humanoid robots. The method performs physically heterogeneity-aware policy decomposition by joint degrees of freedom, conditioning each torque-actuated joint on its own actuation, kinematic, and dynamic attribute descriptors, and through closed-loop policy-actuator co-optimization, jointly adjusts the high-level posture planner and low-level joint actuators using shared interaction trajectories. The authors report that in experiments in MuJoCo and Isaac Gym, H^2MARL consistently outperforms homogeneous, coarse-grained, and independently optimized baseline schemes in locomotion stability, gait coordination, and motion quality. These results are the experimental conclusions of the paper's authors.
AI 根据报道生成 · 6 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROH^2MARL: Hierarchical Heterogeneous Multi-Agent Reinforcement Learning Framework for Humanoid Whole-Body Control
The authors propose H^2MARL, a hierarchical heterogeneous multi-agent reinforcement learning framework for humanoid whole-body control. The method performs policy factorization with physical heterogeneity awareness at the joint degree of freedom (JDoF) level, conditioning each torque-actuated joint on descriptors of its own actuation, kinematic, and dynamic properties, and uses closed-loop policy-executor co-optimization to jointly adapt the high-level pose planner and low-level joint executor with shared interaction trajectories.
Activity for this event
- Current
- 3
- Peak
- 10Nov 5 14:00
- Past 24 hours
- -49%
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.