Robust detection and optimization with decentralized parallel sensor networks

In the first part of this paper, we derive a generic form of optimization for binary hypothesis testing. We show that information theoretical as well as the proposed singular value decomposition (s.v.d.) based optimization methods are special cases of this generalization. In terms of robustness, neither the information theoretical nor the s.v.d. based optimization method has a contribution for decentralized detection. The second part of the paper is concerned with the assignment of the costs for robust distributed detection without a fusion center. We show that the monotonicity rule should remain exactly the same as in distributed detection with a fusion center, meaning that sub-sets of this fusion rule cannot even include the simple fusion rules such as AND or OR.