Akule: Fast and Scalable Multi-Robot Motion Planning via Sparse Interaction Diffusion

Key points

• Diffusion models provide expressive trajectory priors for multi-robot motion planning (MRMP), but scaling them to large teams requires coordinating individual motions without expensive joint computation.1
• This decomposition separates which interactions merit computation from how their effects should be combined into coordinated denoising updates.1
• We train $Rφ$ using matched pairwise trajectory optimizations, then learn $Gω$ and $Uψ$ from physically optimized joint trajectories without requiring a prescribed ground-truth interaction graph.1
• Across diverse MRMP benchmarks, Akule achieves competitive planning success, complete-planner speedups of up to $25\times$, zero-shot population transfer, and successful planning with up to $256$ robots.1

Sources (1)

• [1]Akule: Fast and Scalable Multi-Robot Motion Planning via Sparse Interaction Diffusion
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 07:10 PM

  We introduce Akule, a sparse interaction diffusion framework that augments a frozen single-robot prior with three learned modules: a reusable pairwise correction model $R_φ$, a contextual selector $U_ψ$ that determines which corrections to evaluate, and a composer $G_ω$ that combines their effects through context-dependent signed weights.

  Diffusion models provide expressive trajectory priors for multi-robot motion planning (MRMP), but scaling them to large teams requires coordinating individual motions without expensive joint computation.

Extractive summary: sentences quoted from the sources.