Radar sensor networks with distributed detection capabilities

The goal of this work is to propose a distributed detection method for a network of radars interchanging their measurements through pulse-position modulation. The proposed approach does not need the presence of a fusion center and it is then fully decentralized. The approach is robust against node failures, as the only requirement to get the network gain is that the whole network remains connected, i.e., for every pair of nodes, there is a path, possibly composed of multiple hops, joining them. The decentralized approach is based on a distributed consensus mechanism. However, when conventional consensus algorithms are implemented over realistic channels, the receiver noise gives rise to an error on the final decision statistic that linearly increases with time. We avoid this inconvenient by properly modifying the consensus algorithm to make it suitable for communications over noisy channels. We analyze the performance of the proposed system considering the presence of both observation and communication noise present in the interaction among the radars. In particular, we show that even with a non-coherent integration, the whole system is able to achieve a sensitivity gain equal to the number of radars.