Reproduce “Diffusion for Long-Horizon Multi-Robot Path Planning in Human-Shared Environments” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Vaibhav Sanjay, Yorai Shaoul, Jiaoyang Li
The Multi-Robot Rolling Diffusion (MRRD) framework enables real-time, long-horizon multi-robot path planning in human-shared environments by integrating a rolling-horizon scheme, parallelized Conflict-Based Search (CBS), and urgency-conditioned diffusion. It achieved zero robot-robot collisions and significantly faster planning times (7x faster at 15 robots) compared to baselines, while maintaining human-like trajectories and high success rates in dense crowds and cluttered spaces.
Key insights
- Overcoming the fixed-time constraint in generative planners is crucial for practical robot navigation, achieved by combining a rolling-horizon approach with urgency-based temporal conditioning for dynamic velocity adaptation.
- Parallelizing the initial path generation within a Conflict-Based Search (CBS) framework significantly boosts multi-robot coordination scalability, allowing for real-time performance with large robot teams.
- Differentiated guidance terms in a diffusion model can effectively integrate diverse constraints (robot-robot, robot-human, static obstacle) into a single path generation process, enabling socially aware and collision-free navigation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.