DRL-Based Beamforming Design in RIS-Aided Multi-user Wireless Networks
Keshav K. Singh, Hasan Hasan, Sandeep Kumar Singh, Cunhua Pan, Sudip Biswas · 2023
This paper proposes a deep reinforcement learning (DRL) based passive beamforming design in a reconfigurable intelligent surface (RIS) aided multi-user wireless network. In particular, we consider a RIS to mitigate the effects of blockages and improve the quality of signal transmitted from a base-station to multiple users. Both active and passive beamforming are considered, whereby the active beamformers are formulated through precoder design at the BS, while the proximal policy optimization (PPO) and deep-deterministic policy gradient (DDPG) frameworks based on deep reinforcement learning (DRL) are developed to solve the non-convex problem of optimizing the phase shift (i.e., passive beamforming) of the RIS. Both the proposed DRL based frameworks are compared with each other and with a baseline random phase shift selection algorithm via numerical simulations in terms of the achievable data-rate with respect to the number of RIS elements and transmitted power, that elucidate its feasibility.