Cooperative Multi-agent Constellation Formation Under Sensing and Communication Constraints

Abstract

A group of cooperating vehicles (smart bombs, robots) moves toward a set of (possibly moving) prioritized destinations. During their journey towards the destinations, some of the vehicles may be damaged, but it is also possible that reinforcements may arrive. The objective is to maximize the number of encounters between the vehicles and the high-priority destinations. In this preliminary study, we explore how position sensors and (possibly fading) communication channels can assist the group in performing its task. We show how joint operation, and exchange of observations and estimates, improve convergence. On the other hand unfavorable cooperation attempts can sometimes lead to oscillations and confusion. In unfavorable circumstances our agents suspend cooperation and engage in “every agent for itself” mode. One of the interesting features of the proposed architecture is that individual team members can predict the success or failure of the cooperation mechanism by testing the consistency of assigned target destinations.