Multi-Agent Terraforming: Efficient Multi-Agent Path Finding via Environment Manipulation

David Vainshtain, Oren Salzman
2021 Proceedings of the International Symposium on Combinatorial Search  
Planning collision-free paths for multiple agents operating in close proximity has a myriad of applications ranging from smart warehouses to route planning for airport taxiways. This problem, known as the Multi-Agent Path-Finding (MAPF) problem, is highly relevant to real-world applications in automation and robotics, and has attracted significant research in recent years. While in many applications, the robots are tasked with transporting objects and thus have the means to move obstacles,
more » ... n formulations of the problem prohibit agents from moving obstacles en-route to a task. This often causes agents to take long detours to avoid obstacles instead of simply moving them to clear a path. In this work we present multi-agent terraforming, a novel extension of the MAPF problem that can exploit the fact that the system contains movable obstacles. We build upon leading MAPF solvers and propose an efficient method to solve the multi-agent terraforming problem in a manner that is both complete and optimal. We evaluate our method on scenarios inspired by smart warehouses (such as those of Amazon) and demonstrate that, compared to the classical MAPF formulation, the extra flexibility provided by terraforming facilitates a notable improvement of solution quality.
doi:10.1609/socs.v12i1.18596 fatcat:qwfce4xwq5htnbedax6kzdls4q