Further results on the design of decentralized Bayesian detection systems

Abstract

The decentralized Bayesian detection problem is considered. Some of the design aspects and an analysis of the optimum minimum-cost system are presented. The design of the optimum system is based on an alternate representation of the minimum achievable cost. Using this representation the design of the decentralized Bayesian problem reduces to the optimization of a single function of n variables, thus avoiding the exponential dependence of the cost on the number of variables. The cases of nonidentical as well as identical detectors are treated. Several numerical examples are included