MAPF-World: Action World Model for Multi-Agent Path Finding

TL;DR

MAPF-World introduces an autoregressive action world model for multi-agent path finding that enhances environmental understanding and decision-making, outperforming existing methods in complex scenarios with less data and smaller models.

Contribution

It presents MAPF-World, a novel world model that explicitly captures environmental dynamics and dependencies, improving multi-agent path planning beyond reactive policies.

Findings

Outperforms state-of-the-art learnable MAPF solvers.

Demonstrates superior zero-shot generalization to new scenarios.

Uses significantly less data and smaller models for training.

Abstract

Multi-agent path finding (MAPF) is the problem of planning conflict-free paths from the designated start locations to goal positions for multiple agents. It underlies a variety of real-world tasks, including multi-robot coordination, robot-assisted logistics, and social navigation. Recent decentralized learnable solvers have shown great promise for large-scale MAPF, especially when leveraging foundation models and large datasets. However, these agents are reactive policy models and exhibit limited modeling of environmental temporal dynamics and inter-agent dependencies, resulting in performance degradation in complex, long-term planning scenarios. To address these limitations, we propose MAPF-World, an autoregressive action world model for MAPF that unifies situation understanding and action generation, guiding decisions beyond immediate local observations. It improves situational awareness by explicitly modeling environmental dynamics, including spatial features and temporal dependencies, through future state and actions prediction. By incorporating these predicted futures, MAPF-World enables more informed, coordinated, and far-sighted decision-making, especially in complex multi-agent settings. Furthermore, we augment MAPF benchmarks by introducing an automatic map generator grounded in real-world scenarios, capturing practical map layouts for training and evaluating MAPF solvers. Extensive experiments demonstrate that MAPF-World outperforms state-of-the-art learnable solvers, showcasing superior zero-shot generalization to out-of-distribution cases. Notably, MAPF-World is trained with a 96.5% smaller model size and 92% reduced data.