Multi-agent surveillance and tracking using cyclic stochastic gradient

This paper formulates a solution to a problem in stochastic multi-agent and multi-target surveillance. The research goal is to implement a cyclic stochastic optimization (CSO) application for an ensemble of cooperating, mobile sensing agents executing a surveillance mission. Each agent continually combines information to determine its best estimate of the multi-target system state and then decides on its next action. This decision is selected to optimize a criterion related to the Fisher information matrix. For CSO, only noisy measurements of the objective function are available to each agent, and in this study, each agent attempts to optimize this objective function by calculating its stochastic gradient. This paper examines the implications and applicability of CSO convergence via simulation-based experiments. In these experiments, a cooperating multi-agent ensemble of sensing agents search a region for targets and, using our proposed algorithm, maintain track on all discovered targets.