Optimization of distributed detection networks with tree structures

This paper considers the optimization of tree networks consisting of N nodes under Bayesian criterion. To optimize the system performance, the decision rules at each node must be optimized simultaneously. When the sensors are conditionally independent, it can be shown that the optimal decision rules at each node can be reduced to likelihood ratio tests, and the necessary conditions for the optimum decision thresholds are obtained. Since the parallel and serial detection networks are special cases of tree networks, it is also shown that the results obtained for the tree networks can be easily modified to give the specific results regarding the parallel and serial networks, which agree with the results obtained earlier for these two structures.