Solving Common Goal Conflicts in a Graph-based Multi-agent System

Abstract

We attempt to improve the throughput of graph-based multi-agent AGV system by solving the issue of common goal conflicts. The conflicts are solved by by assigning queue positions based on statistical properties of the graph. These properties include commonly used paths, as well as common fetch and drop locations for orders. These assignments allow agents, which would otherwise remain stationary due to the MAPP algorithm’s inability to find a path to occupied goals, to instead be directed to a suitable intermediate position until the goal becomes available. Keywords: AGV,