AI-Enabled Opportunistic Spectrum Sharing

This chapter considers the AI-enabled opportunistic spectrum sharing between primary users and secondary users in wireless edge networks, where primary users have priorities to use the spectrum and the secondary users can only access the spectrum when the spectrum is sensed to be idle. However, due to the imperfect spectrum sensing, the secondary users may cause interference to primary users. To deal with this issue, this chapter proposes a DQN based modulation and coding scheme (MCS) selection method for primary users to learn the interference pattern and predict the upcoming interference, which can be used to design proper MCS in advance. Simulation results show that, if the MCS switching is ideal, the primary transmission rate of the proposed algorithm is 90 \(\sim 100\%\) of the optimal MCS selection method, which knows interference information at primary users as a prior, and the proposed algorithm is 30 and \(100\%\) better than the upper confidence bandit (UCB) learning algorithm and the signal-to-noise ratio (SNR) based algorithm in terms of primary transmission rate respectively. If the MCS switching cost is considered, the proposed algorithm still outperforms the benchmark algorithms without increasing system overheads.