Trust in multi-agent networks: From self-centered to team-oriented

Abstract

In this work, we incorporate trust into the interaction dynamics for multi-agent systems in order to analyze the effects of trust on human-robot teams. We extend previous work by moving away from a “self-centered” trust model and introduce a so-called “team-oriented” trust model in which each agent increases its trust for its neighbors only if they are collectively contributing to the team's overall goal. The coupling of the trust dynamics and the agents' state dynamics is shown to give rise to an intricate relationship that has the potential to make the team's performance deteriorate under certain circumstances. We derive conditions under which the multi-agent system is guaranteed to achieve a collective objective and provide simulations to corroborate the theoretical findings.