Censoring sensors: A low-communication-rate scheme for distributed detection

We consider a new scheme for distributed detection based on a ''censoring'' or ''send/no-send'' idea, The sensors are assumed to ''censor'' their observations so that each sensor sends to the fusion center only ''informative'' observations, and leaves those deemed ''uninformative'' untransmitted. The main result of this work is that with conditionally independent sensor data and under a communication rate constraint, in order to minimize the probability of error, transmission should occur if and only if the local likelihood ratio value observed by the sensor does not fall in a certain single interval. Similar results are derived from Neyman-Pearson and distance-measure viewpoints. We also discuss simplifications for the most interesting case that the fusion center threshold is high and the communication constraint is severe. We compare censoring with the more common binary-transmission framework and observe its considerable decrease in communication needs, Finally, we explore the use of feedback to achieve optimal performance with very little communication.