RIS-aided transmissions in energy-harvesting cognitive radio networks: A DRL approach

Abstract

Direct transmissions in cognitive radio networks (CRNs) can be easily obstructed by obstacles and channel uncertainty and thus cognitive transmitters normally increase the transmission power to guarantee the quality of service; however, it can deplete limited-capacity batteries and degrade long-term performance. These issues can be solved by reflecting signals to cognitive users (CUs) using a reconfigurable intelligent surface (RIS) and setting appropriate transmission powers. This study investigates RIS-aided downlink of non-orthogonal multiple access (NOMA) CRNs, where RIS can reconstruct transmission environments and a wireless powered-cognitive base station (CBS) opportunistically uses a licensed channel allowing multi-user transmissions in the same frequency and time block. Under stochastic properties of energy harvesting, wireless channels, and primary network behavior, we aim to optimize the assigned power of CUs and RIS phase-shifts jointly to maximize the sum-rate of CRNs. To this end, we formulate an optimization problem as a Markov decision process (MDP) framework. Subsequently, a deep deterministic policy gradient (DDPG) algorithm is adopted to cope with highdimensional continuous states and action spaces in time-varying environments. Simulation results are presented to confirm the superior performance of the proposed scheme over benchmark schemes in which orthogonal multiple access (OMA), long-term, and myopic optimizations are considered.