Channel selection in multi-hop cognitive radio network using reinforcement learning: An experimental study

Abstract

Cognitive radio (CR) is a communication that enables better utilization of radio spectrum. Multi-hop CR network is an emerging area of interest for many researchers in recent years. The concept of multi-hop CR network is widely investigated with many wireless applications, such as video surveillance application. In this paper, we investigate a multi-hop CR network, which can serve as a video surveillance system. Reinforcement learning (RL), which is an artificial intelligence approach, is applied to select and switch to the best possible operating channel. In the CR network context, the unlicensed users (or secondary users, SUs) evade the licensed users' (or primary users', PUs') activities; while in the video surveillance system context, the honest users evade the malicious users' activities (or the denial of service attacks). In our investigation, the multi-hop network comprises of three SUs, namely source, relay and destination nodes. There is a single PU. The experimental setup consists of universal software radio peripheral (USRP) and GNU radio units. Our implementation results show that RL is an effective approach that enables SUs (or honest users) to evade PUs' (or malicious users') activities, and so it improves network performance.