Defensive Configuration for Multi-Agent Perimeter Monitoring

Perimeter monitoring is a required capability of multi-agent systems dealing with security-focused applications. In this paper, we consider the problem of defender configuration for perimeter monitoring, in which a team of defender agents must position and orient themselves to maximize observations of adversarial attackers approaching the perimeter. We envision perimeter monitoring as a crossover of perimeter defense and area coverage. We propose a greedy approximation algorithm that identifies a minimum set of possible fields of view and then selects between them to configure a multi-agent team of defenders by balancing the distance-weighted utility of each observation with the value of covering a diverse set of observations. We evaluate our approach through a variety of realistic multi-agent attack trajectories, showing that our proposed approach achieves near-optimal coverage performance while showing significant gains in time performance.