A Smarter Way to Scale Multi-Agent Pathfinding

This paper formalizes multi-agent pathfinding in N-dimensional grid spaces and introduces LayeredMAPF, a decomposition-based framework that splits large MAPF instances into smaller, solvable subproblems. By leveraging dependency analysis and a connectivity graph, the method maintains completeness while reducing computational complexity. The approach evaluates decomposition quality via maximum subproblem size and empirically analyzes trade-offs between scalability and solution optimality.