Spectrum access in cognitive IoT using reinforcement learning

With the advent of fifth generation technologies for wireless networks and the expansion of the use of the Internet of things, the demands in using spectrum transmission have increased significantly, resulting in a shortage of available spectrum resources to meet these needs. The optimal utilization of the spectrum resources plays an important role to overcome this shortage problem. The Cognitive \(IoT\) (\(CIoT\)) is considered as promising technology to enhance spectrum utilization by accessing the vacant 4G/5G spectrum licensed to a primary user (\(PU\)). The choice between single channel and multiple channels spectrum access is critical in achieving higher data transmission and throughput. In single channel access, the \(CIoT\) waits on the same channel until its availability for usage, while in multiple channels access, \(CIoT\) can switch channels whenever it faces occupied channel, which improves the transmission quality and the achieved throughput. In this paper, a proposed proactive multiple channels spectrum access approach is introduced to enhance the spectrum access of \(CIoT\) through multiple available interfaces, wherein \(CIoT\) utilizes past channel states to predict the forthcoming spectrum availability. The proactive approach uses Reinforcement Learning \((RL)\) algorithm to select the available channels and Bayesian algorithm to predict how long the channel will be unoccupied. The available channels are arranged in descending order of their estimated idle probabilities to enable \(CIoT\) find sufficient idle channels quickly. The \(CIoT\) can use multiple channels simultaneously as long as there are enough free channels for transmission to reduce the spectrum handoffs and the transmission interruptions due to collisions. The sensing accuracy is adapted by achieving a high targeted probability of detection to guarantee primary users protection against harmful interference and lower probability of false alarm to increase the spectrum utilization. The simulation results demonstrate the effectiveness of the proposed approach and show an interesting performance compared with the single channel access model.