Trust networks in multi-robot communities

Abstract

It has been shown that it can be beneficial for robots, which share a common workspace but do not share common goals, to work together and help each other complete their tasks. However, current systems designed to facilitate such actions can be manipulated by dishonest robots seeking to maximize their own benefit. The successful implementation of a system that protects against such activity could benefit groups of robots working near each other, even if the robots are deployed by different and possibly competing organizations. To address this issue, this paper develops and analyzes a distributed trust estimation framework that allows robots to estimate the trustworthiness of other robots in the community and to use this information to determine whether they should cooperate with the same robots. This framework is applied to a specific multi-robot exploration problem, involving the determination of cell types in a discretized workspace. Results show that a team of robots can explore more quickly when other trustworthy robots are present, without sacrificing performance when untrustworthy robots are also present.