Threshold considerations in distributed detection in a network of sensors
Gene T. Whipps, Emre Ertin, Randolph L. Moses · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2012
We consider the problem of distributed sensing and detection using a network of sensor nodes, and the challenges that arise in fusing disparate data. Multiple sensors make local inferences on the state of nature (e.g., the presence of a signal), and those observations are then transmitted to a regional fusion center. The fusion center is tasked to make improved decisions. We develop methods to optimize those decisions. Interoperability between disparate sensor nodes can be addressed by combining similar types of parameters (e.g., direction of arrival and location estimates to better infer location), albeit with varying qualities. As an initial problem, we consider the case where each sensor makes a binary decision on the presence of a signal source and the fusion node combines these to make a more accurate decision. We consider a lossy medium in which signals undergo a range-dependent propagation loss. We determine local thresholds that optimize a performance metric, including both constrained global detection performance and asymptotic error performance. We study the eect sensor node density on detection performance under a network load constraint. The asymptotic performance metrics provide indicators of the amount of value that each sensor contributes to the fusion task.