Robust decentralized estimation fusion in energy-constrained wireless sensor networks with correlated noises

Decentralized estimation fusion based on quantized local estimates is investigated. Motivated by the fact that the variance of quantization noise suffers from bounded uncertainty, we address the robust quantized estimation fusion problem in energy constrained wireless sensor networks. The quantization levels and fusion weights are jointly determined by minimizing worst-case of mean squared error of fused estimate among the uncertainties. Moreover, the effects of observation noise correlations on the fusion performance is explored by considering three different correlation scenarios. The optimal solution of the proposed robust fusion problem in each scenario is derived, in a closed-form or by solving a semidefinite programming. Theoretical analysis indicates that if the correlations exist, the fusion performance might improve or degrade, depending on correlation structure. However, appropriate usage of the correlations in fusion center can asymptotically improve the estimation fusion performance. Numerical simulations are provided to verify the theoretical analysis, and the results illustrate the good performance of the proposed method.