Machine-Learning-Based Optimization for Multiple-IRS-Aided Communication System

Due to the benefits of spectrum and energy efficiency, intelligent reflecting surfaces (IRSs) are regarded as a promising technology for future networks. In this work, we consider single cellular network where multiple IRSs deployed assist downlink transmissions from base station (BS) user equipment (UE). Hence, aim jointly optimize configuration BS active beamforming reflection that meet UE’s QoS while allowing lowest transmit power consumption at BS. Although conventional alternating approach is widely used find converged solutions, its applicability restricted by high complexity, which more severe in dynamic environment. Consequently, an alternative approach, i.e., machine learning (ML), adopted optimal solution with lower complexity. For static UE scenario, propose low-complexity optimization algorithm based on new generalized neural (GRNN). Meanwhile, deep reinforcement (DRL)-based algorithm. Specifically, deterministic policy gradient (DDPG)-based designed address GRNN algorithm’s restrictions efficiently handle scenario. Simulation results confirm proposed algorithms can achieve better power-saving performance convergence noteworthy reduction computation time compared optimization-based approaches. addition, our show total decreases increasing number units IRSs.