Framework for multi-agent path finding A framework for multi-agent path finding with focus on transport systems

Abstract

It is desirable to apply multi agent path finding (MAPF) algorithms to vehicles in transport systems, as it has the potential to reduce transit time and fuel consumption. However, the application is hampered by the high level of abstraction that these problems are typically solved at. This work contributes to the field by introducing a method that lowers the level of abstraction as a post processing phase. We also evaluate a recently developed algorithm that considers mobile agents in continuous time with discrete speeds, and measure its performance when compared to other algorithms. Further, we provide a framework containing these algorithms and the tools required to evaluate and develop them.