Distributed Detection With Adaptive Topology and Nonideal Communication Channels

This paper investigates the increase in detection reliability that an adaptive network (with adaptive topology and channel-dependent decision and fusion rules) can provide, compared with a fixed topology network. We consider a network, consisting of N local detectors and a fusion center (FC), that is tasked with accessing the presence or absence of a phenomenon of interest (POI) at a specific location. The intensity of the signal emitted by the POI when present follows a power law attenuation model. Assuming that sensors transmit binary modulated local decisions over orthogonal nonideal channels, where these channels are modeled as additive white Gaussian noise (AWGN) or fading, we consider two classes of network topologies: 1) sequential topology; and 2) parallel topology. Under Bayesian criterion we derive the optimal channel-dependent rules and error probabilities for these topologies and establish the dependency of the errors on the received signal-to-noise-ratios (SNRs) of the sensing and communication channels. For N=2, our extensive simulation results indicate that the optimality of parallel network is limited and demonstrate the average performance gain of topology adaptation. The results also reveal that the performance gain and the optimal topology are dictated by the system parameters (e.g., transmit powers and distances) and the design search space (i.e., whether the locations of the local detectors are fixed or optimized).