Simultaneous Multi-Robot Motion Planning with Projected Diffusion Models

Jinhao Liang, Jacob K Christopher, Sven Koenig, Ferdinando Fioretto

TL;DR: This paper presents a method combining diffusion models with constrained optimization techniques to generate feasible multi-robot trajectories in complex and highly constrained environments.

Abstract: Recent advances in diffusion models hold significant potential in robotics, enabling the generation of diverse and smooth trajectories directly from raw representations of the environment. Despite this promise, applying diffusion models to motion planning remains challenging due to their difficulty in enforcing critical constraints, such as collision avoidance and kinematic feasibility. These limitations become even more pronounced in Multi-Robot Motion Planning (MRMP), where multiple robots must coordinate in shared spaces. To address these challenges, this work proposes **S**imultaneous **M**RMP **D**iffusion (SMD), a novel approach integrating constrained optimization into the diffusion sampling process to produce collision-free, kinematically feasible trajectories. Additionally, the paper introduces a comprehensive MRMP benchmark to evaluate trajectory planning algorithms across scenarios with varying robot densities, obstacle complexities, and motion constraints. Experimental results show SMD consistently outperforms classical and other learning-based motion planners, achieving higher success rates and efficiency in complex multi-robot environments. The code and implementation are available at https://github.com/RAISELab-atUVA/Diffusion-MRMP.

Lay Summary: Coordinating multiple robots in a shared space is very challenging. Specifically, each robot must avoid collisions and some physical movement rules carefully. Recent advances in AI called diffusion models can generate realistic and flexible paths, but they often struggle to follow these critical rules. Our research presents a new method, named Simultaneous Multi-Robot Motion Planning Diffusion (SMD), that combines diffusion models with optimization techniques to make sure the robots’ paths are feasible, even in complex environments. We also built a new benchmark to test how well different motion planning methods work in various crowded and complex environments. Our experiments show that SMD performs better than traditional and other AI-based planners in these challenging situations.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Primary Area: Deep Learning->Generative Models and Autoencoders

Keywords: Multi-Agent Path Planning, Diffusion Models

Submission Number: 12444