Robust Decentralized Detection in Sensor Networks: Multiple Observations or Multiple Access Points?

This paper presents a general approach to distributed detection in sensor networks with noisy communication links. The sensors are independent and observe a common phenomenon. Communication links are noisy and we model them as binary symmetric channels (BSCs). We first consider a scenario with a single access point (AP), and we follow a Bayesian approach for the evaluation of the probability of decision error. We show that if the noise intensity is above a critical level (i.e., the cross-over probability of the BSC is above a critical value), the probability of decision error at the AP reduces when the AP selectively discards the information transmitted by the sensors with noisy links. Use of multiple observations at the sensors can be traded for increased robustness against channel impairments in the communication links. We then consider the presence of multiple access points and propose a simple architecture to improve the system robustness. Based on the obtained results, we answer the key question in the title