Sequential estimation over noisy channels with distributed node selection

In an autonomous sensor network without a central fusion center, it is desired that any node has the ability to make the final decision once it has enough information about the Phenomenon of Interest (PoI) within a certain confidence level. In this paper, we propose a distributed sequential methodology which updates the current node's estimator based on its own observation and noise corrupted decision from the previous node. We show that sequential processing is useful only when the channel quality of inter-sensor communication links satisfies a certain condition. We develop a distributed node selection algorithm to select the order of processing nodes based on information utilities and the communication cost. In the proposed scheme, each node only needs to keep track of its neighbor nodes leading to reduced complexity. Simulation results show that a significant reduction in the required number of processing nodes to achieve a desired performance level is obtained compared to that with nearest node selection method.