Deep reinforcement learning based power minimization for RIS-assisted MISO-OFDM systems

In this paper, we investigate the downlink orthogonal frequency division multiplexing (OFDM) transmission system assisted by reconfigurable intelligent surfaces (RISs). Considering multiple antennas at the base station (BS) and multiple single-antenna users, the joint optimization of precoder at the BS and the phase shift design at the RIS is studied to minimize the transmit power under the constraint of the certain quality-of-service. A deep reinforcement learning (DRL) based algorithm is proposed, in which maximum ratio transmission (MRT) precoding is utilized at the BS and the twin delayed deep deterministic policy gradient (TD3) method is utilized for RIS phase shift optimization. Numerical results demonstrate that the proposed DRL based algorithm can achieve a transmit power almost the same with the lower bound achieved by manifold optimization (MO) algorithm while has much less computation delay.