Research map / Image generation / Autoregressive and latent generation
Simulation-Informed Diffusion for Decentralized Multi-robot Motion Planning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Simulation-Informed Diffusion (SID), where a constraint-aware diffusion model both simulates neighbors' future trajectories and plans each robot's path, with conflict-triggered minimal communication for decentralized multi-robot motion planning.
Problem
Decentralized multi-robot motion planning from local observations with limited communication; snapshot-based planners cannot anticipate neighbors' future behavior, especially in dense or cluttered environments.
Approach
Use one constraint-aware diffusion model (diffusion sampling with projection onto collision and kinematic constraints via augmented Lagrangian) as both a neighbor-trajectory simulator and an ego planner, and communicate only when local planning becomes infeasible, with priority ordering by distance to goal.
Key results
- SID achieves 100% success on basic and dense maps with up to 9 robots, and 84% and 96% on room and shelf maps with 9 robots, versus 20% and 12% for ORCA.
- On extended maps with 18 robots, SID attains 80% (basic) and 88% (dense) success while ORCA reaches 0% and 4%; IA-MPC and MIMIC-D solve no extended cases.
- SID's communication frequency stays low (at most 0.41 on extended maps) compared with a fixed 8 for IA-MPC.
Surprising findings
- MIMIC-D, a diffusion-policy baseline, fails on most nine-robot settings despite being generative, because its generated trajectories do not satisfy planning constraints.
Limitations
- Evaluated only in 2D with disk robots and static obstacles.
- Constraints must be explicitly specified; implicit constraints left to future work.
- Higher trajectory acceleration and longer running times than ORCA; smoothness values reported only on successful instances.
Experimental setup
Datasets: modified MRMP benchmark (basic, dense, room, shelf maps)
Benchmarks: modified MRMP benchmark (basic, dense, room, shelf) extended basic and dense maps up to 18 robots challenging 108-robot scenario
Metrics: success rate communication frequency smoothness travel time running time
Baselines: ORCA IA-MPC MIMIC-D
Builds on: projected diffusion for motion planning score-based diffusion models space-time A* MIMIC-D
Closest papers by meaning
- Multi-Robot Motion Planning with Diffusion Models (ICLR 2025 · similarity 0.91)
- Simultaneous Multi-Robot Motion Planning with Projected Diffusion Models (ICML 2025 · similarity 0.88)
- Constrained Diffusers for Safe Planning and Control (NeurIPS 2025 · similarity 0.86)
- SafeDiffuser: Safe Planning with Diffusion Probabilistic Models (ICLR 2025 · similarity 0.86)
- Model-based Diffusion for Trajectory Optimization (NeurIPS 2024 · similarity 0.86)
- DiRecT: Safe Diffusion-Based Planning via Receding-Horizon Denoising (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.