Scalable Decentralized Unlabeled Motion Planning using Graph Neural Networks

A decentralized graph neural network policy can effectively solve the unlabeled motion planning problem, outperforming centralized and greedy decentralized approaches while scaling to large multi-robot systems.