Collaborative Filtering for Coordinated Monitoring in Sensor Networks

Wireless sensor network technology enables large-scale monitoring of the environment and new types of data-intensive applications. Sensor nodes can be deployed to monitor the environment in a coordinated fashion. The information content in such a scenario is reflected by a bulk of measurements, not only by individual measurements. For coordinated acquisition and refinement of information, dependencies between sensor measurements need to be taken into account. We formulate the problem of coordinated monitoring in the sensor network as a problem of collaborative filtering (CF), and present our solution in a structural health monitoring application, that is, monitoring of man-made structures for diagnostic purposes. In our setting, we select information for the novelty detection problem by rating signal features locally on individual sensors and use CF to infer globally appropriate selection of features to monitor. We empirically test the methodology on accelerometer data measured from two physical model structures.