Abstract—Decentralized sensor networks are collectionsof individual local sensors that observe a common phenomenon,quantize their observations, and send this quantizedinformation to a central processor (fusion center)which then makes a global decision about the phenomenon.Most of the existing literature in this field consider onlythe data fusion aspect of this problem, i.e., the statisticalhypothesis testing and optimal combining of the informationobtained by the local sensors. In this paper, we proposea Parallel Genetic Algorithm (PGA) for optimizing theprobability of global detection error performance of aparallel decentralized sensor network. Specifically, we usethe PGA to simultaneously optimize both the fusion ruleand the local decision rules. We show that our approachprovides results comparable to those obtained by using aGA and gradient-based algorithm from previous work byAldosari and Moura, with reduced complexity. We considerboth the cases of identical (homogeneous) and non-identical(heterogeneous) sensors and demonstrate that our algorithmconverges to the same optimal solution in both cases. We alsodiscuss the effect of the quality of the initial solution on theconvergence of the PGA.

Index Terms—Decentralized sensor networks, distributeddetection, optimal fusion rule, genetic algorithms

Cite: Nithya Gnanapandithan, Balasubramaniam Natarajan, "Joint Optimization of Local and Fusion Rules in a Decentralized Sensor Network," Journal of Communications, vol. 1, no. 6, pp. 9-17, 2006.