Affective recruitment of distributed heterogeneous agents

Abstract

Members of multi-robot teams may need to collaborate to accomplish a task due to differences in capabilities. This paper describes an extension of the ALLIANCE architecture that enables agent recruitment within a decentralized UAV-UGV robot team without task preemption but 1) uses a formal model of emotions and 2) handles heterogeneity. Affective computing allows recruitment to be robust under loss of communication between agents and minimizes the number of messages passed. Data from 66 simulations show that the affective strategy succeeds with a random message loss rate up to 25% and requires 19.1% fewer messages to be sent compared to greedy and random, and that of these, affective scales best with team size. Comparisons of broadcast to unicast messaging are also made in simulation.