Event
HA-DiPO Heterogeneous Robot Collaborative Transport Framework Released
1 reports · 1 sources
Xuebin Li and other authors submitted a paper proposing HA-DiPO—a hierarchical multi-agent reinforcement learning framework operating under centralized training and decentralized execution (CTDE) for heterogeneous legged and wheeled-legged robots collaboratively transporting an unsecured shared payload. The framework's high level uses a centralized critic to jointly train each robot's independent coordination policy, representing multimodal coordination responses with a diffusion policy; the low level consists of each embodiment's proprioceptive locomotion policy tracking the coordination commands. The authors state that the coordination policy is learned from scratch through reward-driven simulation interaction without demonstrations; at execution time, each robot generates commands using only local robot and payload observations, without relying on partner states or inter-robot communication. The authors state that real-world experiments demonstrated quadruped-wheeled-legged robot and human-robot collaborative transport.
Report timeline
- HA-DiPO: A Diffusion Policy Optimization Framework for Heterogeneous Robot Cooperative Transport
The authors propose HA-DiPO, a hierarchical multi-agent reinforcement learning framework operating under centralized training with decentralized execution (CTDE) for heterogeneous legged robots cooperatively transporting an unfixed shared payload. At the high level, a centralized critic jointly trains each robot's independent coordination policy, and diffusion policies represent multimodal coordination responses; at the low level, embodiment-specific locomotion policies track the coordination commands. The policies are learned from scratch through reward-driven simulation interaction, and at execution they use only local observations with no inter-robot communication.
arXiv cs.RO ·