Distributed Compression-Estimation Using Wireless Sensor Networks

In this paper we consider deterministic parameter estimation problems. We study the intertwining of quantization and estimation in general and shows particular results in 1 low SNR situations where the noise standard deviation is in the order of the parameters dynamic range and 2 universal estimation when the sensor data and noise model are unknown. The goal is to understand how the signal processing capability of a WSN scales up with its size, and to develop robust distributed signal processing algorithms and protocols with low bandwidth requirement and optimal performance. We show that for universal estimation in low signal to noise ratio SNR, the universal distributed estimators not only exist but achieve performance close to that of estimators based on the original un-quantized observations. We also generalize these results a Bayesian estimation framework with a particular application to state estimation of dynamic stochastic processes.