Cooperative Spectrum Sensing for Cognitive Radios: Bounds and Algorithms

This paper considers wireless spectrum sensing in harsh environments dominated by shadowing and fading. By modeling the network of the secondary users as Markov random fields and pulling a group of secondary users to cooperate through distributed probabilistic inference, effective sensing and fusion can be achieved. The proposed framework subsumes belief propagation, as well as conventional weighted hard/soft combining (such as maximal ratio combining and equal gain combing). It can also account for the distance-dependent correlation among individual sensing results by setting appropriate compatibility function. Theoretic upper and lower bounds are derived, demonstrating the significant gains made possible by effective cooperation. Extensive simulations confirm the analytical results.