Distributed energy-efficient estimation in spatially correlated wireless sensor networks

We study the problem of maximizing lifetime in sensor networks that are deployed for estimating an unknown parameter or process. Sensors take measurements and transport them in multi-hop fashion to a fusion center (FC) for Maximum Likelihood (ML) estimation of this process. A prime distinguishing feature of wireless sensor networks is the spatial correlation among measurements of nearby sensors. We investigate the way spatial correlation shapes the tradeoff between estimation quality and energy efficiency, according to which more measurement data improve estimation quality, but they are more energy-costly to transport. If the effect of spatial correlation is understood, then a given estimation quality can be achieved with minimum data redundancy and energy consumption. We study the dynamic control of the sensor sampling rates and the routes to the FC. Sensor attributes such as spatial correlation, measurement accuracy and energy reserve as well as the quality of wireless links, collectively affect the sampling rates and routes to the FC. Further, due to spatial correlation, sensor sampling rates alter the joint probability density functions (p.d.f’s) for sensor readings and thus they affect the average estimation error. We show that the optimization problem can be decomposed into smaller ones, where each sensor autonomously takes its sampling rate and next-hop forwarding decisions, and we propose an iterative, low-overhead primal–dual algorithm. Our work yields interesting insights on the fundamental tradeoff between network lifetime and estimation quality, it provides a clear intuition on when spatial correlation is beneficial for the tradeoff above, and it provides a solution with distributed sensor coordination.