Cost, Precision, and Task Structure in Aggression-Based Arbitration for Minimalist Robot Cooperation

Abstract

This paper reexamines a multi-robot transportation task, introduced and studied by Vaughan and his collaborators, in which constrained space induces inter-agent interference. Previous research demonstrated the effectiveness of an arbitration mechanism inspired by biological signaling where the level of aggression displayed by each agent effectively prioritizes the limited resources. This paper shows that apart from determining the correct fitness of an individual several other factors, such as signaling cost, precision of the outcome and properties of the resource and task are key to determine an effective arbitration technique. Based on these factors we present a taxonomy of the arbitration mechanisms.

The large signalling costs incurred by our simple robots using minimal set of sensors permit us to identify scenarios in which a dominance hierarchy outperforms, not only to no arbitration, but also aggression-based mechanisms. We identify how memory of past interactions can be used to the advantage of an agent, albeit with a trade-off between cost and outcome accuracy. We also show that the importance of a particular aggressive interaction to long-term task performance is not trivial to determine and depends on the task structure. Results help us identify instances where agents may manipulate interactions to alter the frequency and duration of aggressive encounters, affecting the overall task performance.