Multi-Agent Path Finding with Prioritized Communication Learning

TL;DR

This paper introduces PICO, a prioritized communication learning method for multi-agent pathfinding that improves success and collision rates in large-scale tasks by incorporating implicit planning priorities into decentralized reinforcement learning.

Contribution

The paper presents a novel prioritized communication learning approach that integrates implicit planning priorities into decentralized multi-agent reinforcement learning for MAPF.

Findings

PICO outperforms state-of-the-art learning-based planners in success rates.

PICO significantly reduces collision rates in large-scale MAPF tasks.

The method effectively incorporates implicit priorities to improve decentralized planning.

Abstract

Multi-agent pathfinding (MAPF) has been widely used to solve large-scale real-world problems, e.g., automation warehouses. The learning-based, fully decentralized framework has been introduced to alleviate real-time problems and simultaneously pursue optimal planning policy. However, existing methods might generate significantly more vertex conflicts (or collisions), which lead to a low success rate or more makespan. In this paper, we propose a PrIoritized COmmunication learning method (PICO), which incorporates the \textit{implicit} planning priorities into the communication topology within the decentralized multi-agent reinforcement learning framework. Assembling with the classic coupled planners, the implicit priority learning module can be utilized to form the dynamic communication topology, which also builds an effective collision-avoiding mechanism. PICO performs significantly better in large-scale MAPF tasks in success rates and collision rates than state-of-the-art learning-based planners.