Spectrum sensing utilizing unmanned aerial vehicles (UAVs) has become increasingly popular due to their advantageous line of sight (LoS) communication links. In this study, we propose and evaluate an adaptive spectrum sensing scheme with reinforcement learning for UAV based cognitive radio systems. Instead of employing multiple secondary users (SUs) as in a terrestrial cooperative spectrum sensing setup with a fusion center (FC), our approach involves a single UAV performing virtual cooperative sensing by flying on a circular aerial trajectory. The UAV's sensing period consists of virtual mini-sensing slots akin to a group of SUs. The UAV enhances sensing reliability by performing local spectrum sensing within each mini-slot and combines the collected data using the voting scheme to make collective decisions. Moreover, traditional UAV-based virtual cooperative sensing schemes are facing problems in adjusting the UAV velocity, radius, and flight height to get maximum network throughput. Therefore, these variables are manually tuned for the UAV and are kept fixed. However, in the proposed scheme UAV follows reinforcement learning to adjust the UAV parameters by increasing and decreasing their levels intelligently to obtain higher rewards as the network throughput.
This research outcome is helped in part by the National Research Foundation of Korea (NRF) grants financed by the Korea government (MSIT) (No. 2021R1A2C1013150 and 2022R1F1A1074556).