Deep-Q-Network Based Sum Rate Maximization for IRS-Assisted NOMA Communication Systems

Intelligent Reflecting Surfaces (IRSs) have emerged as a promising technology for enhancing the performance of next generation wireless communication networks by intelligently controlling the wireless propagation environment. A critical challenge in IRS-assisted systems is the optimization of the phase shift angles of individual IRS element to maximize signal quality and network efficiency. This paper presents a reinforcement learning-based approach, utilizing a Deep Q-Network (DQN), to optimize the phase shift of IRS elements dynamically in a wireless communication system model. The proposed method formulates the IRS phase shift design as a Markov Decision Process (MDP), where the agent learns to adjust phase shifts to maximize the received signal-to-noise ratio (SNR) and achievable sum data rate. Simulation results demonstrate that the DQN algorithm can effectively learn optimal phase shift policies under varying channel conditions, outperforming traditional optimization techniques in terms of computational efficiency. In this paper, we consider implementing a system model consisting of IRS aided (NOMA) communication system to maximize the proposed system achievable sum rate. The proposed simulation results obtained by running the proposed DQN algorithm on Matlab tools, the obtained results show the effectiveness of the proposed DQN reinforcement learning algorithm in solving the proposed phase shift of IRS elements optimization problem. Also, a comparison between the proposed DQN algorithm and other algorithms has been introduced in the simulation results to show the superior effect of the DQN algorithm on solving the proposed optimization problem and achieving maximum sum data rate for the communication system compared with other algorithms.