The impact of fading on decentralized detection in power constrained wireless sensor networks

We study a binary decentralized detection problem in which a set of sensor nodes provides partial information about the state of nature to a fusion center. Sensor nodes have access to conditionally independent observations, given the state of nature, and they transmit their data over separate wireless channels. The communication link between each node and the fusion center is subject to fading, with certain nodes possibly having much better connections than others. Upon reception of the information, the fusion center attempts to accurately reconstruct the state of nature. Large deviation theory is employed to obtain design guidelines for wireless sensor networks with a large number of nodes. The normalized Chernoff information is shown to be an appropriate performance metric to compare prospective sensor nodes. For the specific example of binary sensor nodes sending data over Rayleigh fading channels, the performance loss due to fading is found to be small.