Multi-Agent Path Finding for Non-Conflicting Paths: An Infinite Speed Conflict-Based Search Approach
Yaodong Hou, Gang Tao, Zheng Zang, Guojie Kong, Jianwei Gong · 2023
Multi-Agent Path Finding (MAPF) is the problem of finding a set of non-conflicting paths for multiple agents on a graph, which is NP-hard to solve optimally. This paper proposes an infinite speed Conflict-Based Search (IS-CBS) algorithm to solve the MAPF problem when the agents move at different speeds and the edges correspond to roads of varying lengths. In addition, three distinct approaches are introduced to calculate admissible heuristics for IS-CBS: the sum approach which uses the sum of all heuristics as the final heuristic, the Pareto approach which builds a Pareto set and selects nodes from it, and the alternation approach which alternates between heuristics in search iterations. In our experiment, IS-CBS with heuristics outperforms IS-CBS in terms of success rate, runtime, and number of expanded nodes which is reduced by up to a factor of 21.