Decentralised sensor network performance with correlated observations

Abstract

In this paper, we use a novel Parallel Genetic Algorithm (PGA) approach to simultaneously optimise both the fusion rule and the local decision rules of a decentralised sensor network with correlated observations. The optimisation is performed with respect to the probability of error at the fusion centre. We show that our algorithm converges to a majority-like fusion rule irrespective of the degree of correlation and that the local decision rules play a key role in determining the performance of the overall system. By fixing the fusion rule and optimising only the local rules, we demonstrate that systems having different fusion rules can all provide similar performance if the local rules are chosen appropriately. We also show that the performance of the system degrades with increase in the correlation between the observations.