Power allocation for power-limited sensor networks with application in object classification

This publication analyzes the power allocation problem for a distributed wireless sensor network which is based on ultra-wide bandwidth communication technology. The network has power-limited sensor nodes and it is used to classify target objects. In the considered scenarios, the absence, the presence, or the type of an object is observed by the sensors independently. Since the observations are transmitted over noisy communication channels, and are thus unreliable, the disturbed observations are fused into a reliable global decision in order to increase the overall classification probability. In [1] an information theoretic approach, that aims at maximization of the mutual information, has been employed. It enables the analytical allocation of the given total power to the sensor nodes so as to optimize the overall classification probability. We follow the same idea and improve on the results in [1] by a smart selection of the sensor nodes. Furthermore, we investigate the power constraint per sensor node and extend the results hereby.