Abstract
Data gathering is a basic activity of many wireless sen- sor network applications. We propose a novel collaborative data gathering approach utilizing data co?occurrence, which is different from data correlation. Our approach of- fers a trade-off between communication costs of data gathering versus errors at estimating the sensor measurements at the base station, by having sensors with co?occurring measurements alternate in transmitting such measurements to the base station, and having the base station make infer- ences about sensor measurements utilizing only the trans- mitted data. We describe two effective methods for in- network detecting measurements co-occurrence among sen- sors, an efficient protocol for scheduling the transmission of measurements, and a simple algorithm for measurement inference. Our simulation results on synthetic and real datasets show a substantial (up to 65%) reduction on the communication costs of data gathering with few number of inference errors at the base station.