RIS Assisted Millimeter Wave Communication with Imperfect CSI Based on DRL

In order to improve the data rate received at the user's end, Reconfigurable Intelligent Surface (RIS), an artificial material that can control electromagnetic wave characteristics, can reshape the wireless channel. This allows it to be built in millimeter-wave non-line-of-sight links that are obscured by obstacles. This paper combines a long-short-term memory (LSTM) network with depth-deterministic policy gradient (DDPG), taking into account the fact that imperfect channel state information (CSI) can also result in a degradation of communication system performance. By using the LSTM neural network's memory function to control data delivery, the combined approach optimizes both the RIS reflective phase shift and the base station transmitting beam-forming matrix in an undesirable environment. The simulation findings demonstrate that, in comparison to the current deep reinforcement learning algorithms, the LSTM-DDPG algorithm can offer customers more stable and efficient data rates.