Disruption of Connectivity Graphs in Uncertain Multi-Agent Systems

Brian Reily, Caden Coniff, John G. Rogers, Christopher M. Reardon

Abstract: Multi-agent systems have become ever-present in modern society, whether as multi-robot teams, sensor networks, or social networks. While ensuring the connectivity and robustness of multi-agent systems has seen extensive research, the problem of disrupting the connectivity of a multi-agent system has remained largely unaddressed. Yet, this capability can be essential in certain applications, such as responding to a hostile multi-robot system or controlling the flow of disinformation in a social network. In this paper we propose a novel method to disrupt the connectivity of a multi-agent system with uncertain relationships. We represent a multi-agent system as a graph, with edges denoting the probability of communication between agents. We introduce the problem of identifying a subgraph which minimizes the overall connectivity of the multi-agent system. We formulate a novel approach to identify optimal sets of vertices to remove by approximating a minimization of the algebraic connectivity, given constraints on the number of vertices to disconnect. We show through evaluation on simulated multi-agent systems that our approach is able to effectively disrupt the connectivity of a multi-agent system, and discuss its comparative complexity to existing approaches while attaining these superior results.