This paper introduces Stein Variational Belief Propagation (SVBP), a new algorithm for inference over nonparametric marginal distributions of nodes in a graph. This is applied to multi-robot coordination, modeling a robot swarm as a graphical model and performing inference for each robot. The experiments show that SVBP represents multi-modal distributions better than sampling-based or Gaussian baselines, improving performance on perception and planning tasks. SVBP’s ability to represent diverse trajectories for decentralized multi-robot planning makes it less prone to deadlock scenarios.
Publication date: 29 Nov 2023
Project Page: unknown
Paper: https://arxiv.org/pdf/2311.16916
