Distributed Sensing of Interference Pattern in Dense Cooperative Wireless Networks

In this paper we consider the problem of distributed spectrum sensing in multiple selforganizing networks sharing the same timefrequency resources. Each of the networks allocates autonomously radio resources so as to minimize mutual interference. Interference sensing is part of this cognitive framework where sensing devices, or secondary users (SUs), exchange local estimates to cooperatively recognize and track the overall time-varying interference patterns caused by primary users (PUs). PUs are assumed to perform periodic transmission over pre-defined (but unknown to SUs) time-frequency hopped resources. Detection by the SUs is based on local processing and iterated exchanges of local decision with neighbors, so as to enable global fusion of sensed data as for an equivalent centralized approach. We propose a weighted-average consensus algorithm nested within a decision-directed procedure for distributed Bayesian detection of the PU spectrum occupancy. The distributed approach provides the estimate of the complete interference pattern to each SU regardless of the incomplete visibility at each node. Performance analysis is carried out both on simulated and real scenarios with mixed coexisting WiFi and ZigBee devices.