The collaborative iterative search approach to multi agent path finding

Abstract

This thesis presents a new approach to obtaining optimal and complete solutions to Multi Agent Path Finding (MAPF) problems called Collaborative Iterative Search (CIS). CIS employs a conflict based scheme inspired by the Conflict Based Search (CBS) algorithm and extends this to include a linear order lower level search. The structure of Planar Graphs is leveraged, permitting further optimization of the algorithm. This takes the form of reasoning-based culling of the search space, while maintaining optimality and completeness. Benchmarks provided demonstrate significant performance gains over the existing state of the art, particularly in the case of sparsely populated maps. The thesis draws to a conclusion with a summary of proposed future work.