Solving Multi-Agent Sokoban via LaCAM

2026-09-30

AI Summary

This study addresses the challenges of branching factor explosion and the joint difficulty of task assignment and path coordination in multi-agent Sokoban planning. We propose a scalable LaCAM-based planner that leverages Multi-Agent Path Finding (MAPF) as a core primitive. By integrating a centralized task assignment mechanism, the approach exploits recent advances in MAPF to effectively compress the search space, enabling efficient resolution of large-scale instances. Experimental results demonstrate that the proposed planner can efficiently handle complex scenarios involving dozens of agents while strictly preserving algorithmic completeness and asymptotic optimality guarantees. This work establishes a novel paradigm for multi-agent cooperative planning that combines rigorous theoretical assurances with practical scalability.

Abstract

Sokoban, a puzzle game in which an agent pushes boxes onto unlabelled target locations in a grid world, is a long-standing benchmark planning problem. While it is easy to see the connection to practical applications such as warehouse logistics with autonomous forklifts, its multi-agent counterpart has remained underdeveloped. This is because Multi-Agent Sokoban is substantially more difficult due to factors specific to multi-agent planning, such as the rapidly growing branching factor as the number of agents grows and the need to handle integrated task assignment and collision-free pathfinding. In this paper, we show that a scalable planner for Multi-Agent Sokoban can be designed by leveraging recent advances in multi-agent pathfinding (MAPF). Specifically, our Sokoban-LaCAM efficiently solves instances involving tens of agents and boxes while preserving both completeness and eventual optimality guarantees. This provides evidence that MAPF can serve as a powerful primitive for solving broader collective automation problems.

Problem

• Multi-Agent Sokoban
• Multi-Agent Pathfinding
• Task Assignment
• Collision-Free Pathfinding
• Planning

Innovation

• Multi-Agent Sokoban
• LaCAM
• Multi-Agent Pathfinding (MAPF)
• Task Assignment
• Completeness and Optimality