Simulation-Informed Diffusion for Decentralized Multi-robot Motion Planning
A new framework called Simulation-Informed Diffusion (SID) addresses the challenge of decentralized multi-robot motion planning. Existing planners generate trajectories from static local observations, failing to anticipate neighboring robots' future behavior, especially in cluttered environments with many robots. SID uses constraint-aware diffusion models (CADM) to first simulate neighboring robots' future trajectories from their current states, then plans each robot's own trajectory under safety constraints informed by those simulations. The framework operates without global sensing or reliable communication, relying only on local observations. The paper is published on arXiv with ID 2605.27697.
Key facts
- SID is a decentralized framework for multi-robot motion planning.
- It uses constraint-aware diffusion models (CADM).
- SID simulates future trajectories of neighboring robots from local observations.
- It plans each robot's own trajectory under safety constraints.
- The approach does not require global sensing or reliable communication.
- Existing planners generate trajectories from static snapshots.
- The limitation is critical with increasing robot numbers and clutter.
- The paper is available on arXiv:2605.27697.
Entities
Institutions
- arXiv