A Spectral Clustering Approach to Validating Sensors via Their Peers in Distributed Sensor Networks

In a distributed sensor network, the goodness of a sensor may change according to its current device status (e.g., health of hardware) and environment (e.g., wireless reception conditions at the sensor location). As a result, it is often necessary to validate periodically sensors in the field, in order to identify those which no longer work properly and eliminate them from applications' use. In this paper, we describe a spectral clustering approach of using peer sensors to identify these bad sensors. Using a simple model problem, we describe how our sensor validation method works and demonstrate its performance in simulation.