Asymptotic fusion performance in a power constrained, distributed wireless sensor network

Abstract

The performance of decentralized detection in power constrained wireless sensor network is analyzed. We consider two distributed processing schemes; the analog relay amplifier local processing and the quantized decision local processing. In the first model, the system is assumed to be subjected to a total power constraint and the distributed nodes are assumed to perform analog relay amplifier local processing. Under these assumptions, we show that, we can monotonically improve the fusion performance by using non-orthogonal sensor-to-fusion center communication, as opposed to orthogonal communication. In order to quantify the performance, we employ circulant matrix theory to derive closed form asymptotic expressions. These asymptotic results allow us to observe the effect of each parameter on the total system performance. Numerical results show that these asymptotics provide a very good approximation to the exact performance, even for a small number of sensors. In the second model, distributed nodes are assumed to make binary decisions, in an inhomogeneous sensor system. i.e. the channel statistics are not identical among the sensors. We derive the asymptotic fusion performance via the Lindberg-Feller central limit theorem. A power allocation scheme is also proposed based on the well-known water-filling algorithm. The proposed scheme shows better performance corresponds to the equal power assignment when the total power is sufficient with respect to the network size. However, as we increase the network size the water-filling gain diminishes.