Hybrid MRT and ZF Learning for Energy-Efficient Transmission in Multi-RIS-Assisted Networks

The deep reinforcement learning-based transmission design for reconfigurable intelligent surface (RIS) assisted multiple input single output networks is investigated in this research. An energy efficiency (EE) maximization problem is formulated under constraints of the rate requirement, the power budget and the phase shift coefficient. To be adaptable to various wireless channel conditions, a novel model-based beamforming design, namely the hybrid maximum ratio transmission (MRT) and zero-forcing (ZF) scheme, is proposed in the action space. Besides, a new activation function is adopted to handle the power budget constraint. The proximal policy optimization (PPO) approach is adopted to learn the optimal policy. Numerical results validate the efficacy of the hybrid MRT and ZF scheme as well as the PPO-based algorithm.