Reinforcement Learning Based Decision Fusion Scheme for Cooperative Spectrum Sensing in Cognitive Radios

Spectrum Sensing is a key module in Cognitive Radios (CR) for detecting spectrum holes. The performance of spectrum sensing algorithms is compromised due to channel impairments, such as, multi-path fading and shadowing. Cooperative Spectrum Sensing (CSS) scheme mitigates the above issues and improves the spatial diversity gain of Secondary Users (SUs). In this paper, we present Reinforcement Learning (RL) based CSS scheme with the objective of improving cooperative sensing accuracy by maximizing expected cumulative reward. Using reinforcement learning, the Fusion Center(FC) makes a global decision by interacting with the radio environment which consists of cooperative SUs and primary transmitter. The cooperative SUs are deployed randomly in a fading wireless channel environment modeled as a Markov Decision Process (MDP). The optimal solution of RL based CSS algorithm is formulated using policy iteration to meet the requirements of IEEE 802.22 Wireless Regional Area Network (WRAN) standard. The simulation results show that the RL based CSS scheme improves the detection performance under channel fading/shadowing and overall cooperative learning capability.