Hoplites: A Market Framework for Complex Tight Coordination in MultiAgent Teams

摘要

In this paper we present a new class of tasks for multi-robot teams: those that require constant complex interaction between teammates. Much research has been done in the area of multi-robot coordination, but no existing framework meets the technical demands of such tasks. We have developed Hoplites in response to the need for a more capable framework. Hoplites is a market-based framework that couples planning with both passive and active coordination strategies. It enables robots to change coordination strategies as the needs of the task change. Further, it efficiently facilitates tight coordination between multiple robots. We compare the performances of Hoplites and existing coordination frameworks in a security sweep domain. Our results show that Hoplites significantly improves the quality of solutions found by the team, particularly in the most complex instances of the domain. I