Design of Distributed Detection Systems with Correlated Heterogeneous Sensors

The optimization of sensor thresholds and fusion rule for heterogeneous and correlated sensor suite is accomplished through a particle swarm optimization algorithm. Different correlation structures are assumed and the effect of correlation on the choice of final fusion rule and thresholds is analyzed. Optimal decision fusion for correlated sensors includes estimation of 2 joint probabilities. Bahadur-Lazarfeld expansion is used to reduce the computational burden. Bahadur-lazarfeld expansion reduces the burden to evaluation of N joint probabilities. Numerical integration is proposed to evaluate the joint probability. Examples using Gaussian distributions assumed under both the hypothesis are presented. The results achieved using a particle swarm optimization algorithm are compared to the traditional decision fusion strategies based on the fusion methodology developed by Moshe Kam et al.