Reinforcement learning approach for centralized Cognitive Radio systems

Providing that licensed or Primary Users (PUs) are oblivious to the presence of unlicensed or Secondary Users (SUs), Cognitive Radio (CR) enables the SUs to use underutilized licensed spectrum (or white spaces) opportunistically and temporarily. A centralized CR system is an architectural model for a wide range of applications for example wireless medical telemetry service and medical implant communications service. As an enabling technology for white space exploitation, context awareness and intelligence (or cognition cycle, CC) remains the key characteristics of CR for using the underutilized licensed spectrum in an efficient manner. In this paper, we provide investigation into the application of a stateful Reinforcement Learning (RL) approach, to realize the conceptual CC in centralized static and mobile networks in the presence of many PUs. We investigate the use of RL with respect to Dynamic Channel Selection (DCS) that helps the SU Base Station (BS) to select channels adaptively for data transmission between different SU hosts. The purpose is to enhance the Quality of Service (QoS), particularly to maximise throughput and reduce delay by means of minimizing the number of channel switches. Simulation results reveal that RL achieves good performance and that the learning and exploration characteristics should converge to a low value to optimise performance.