Accelerating Search-Based Planning for Multi-Robot Manipulation by Leveraging Online-Generated Experiences

We present a novel framework that integrates diffusion models with search-based planning algorithms to enable large robot teams to operate under learned motion patterns while only relying single-robot data.