Bewertung von Multi-Agent Path Finding Verfahren für deren Einsatz in robotisierten Logistiksystemen

Abstract

In recent time, logistics systems have been in place in which hundreds of autonomous mobile robots are used for tasks such as picking or sorting. Path planning for the robots navigating in these systems is still a relevant problem, for which Multi-Agent Path Finding (MAPF) methods have been considered for some time. However, there has been no consensus on which evaluation criteria allow for estimating the suitability of a MAPF method for a particular application. In this work, we therefore derive qualitative evaluation criteria (scalability, conflict management, and solution quality), that enable such a suitability assessment. We use these criteria to find that Priority Based Search is suitable for Robotic Mobile Fulfillment systems, while Explicit Estimation Conflict Based Search is appropriate in robotic sorting systems. For robotic-based production logistics, we recommend the Conflict Based Search algorithm.