Fairness in Reconfigurable Intelligent Surface (RIS) assisted MU-MISO systems Using DRL

This work uses Deep Reinforcement Learning framework to maximize downlink sum spectral efficiency of Reconfigurable Intelligent Surface (RIS) Assisted Multi-User Multiple Input Single Output system. Controller at RIS runs the DRL algorithm. Controller is assumed to have perfect knowledge of channel between Base station and RIS as well as of the channel between RIS and multiple single antenna users. The algorithm determines optimal beamforming gains and phase angles such that equal data rates are allocated to all users, along with maximizing the sum rate. This work thereby promotes equitable resource access and enhances overall user satisfaction. Sum rate as a function of maximum allowed transmit power at base station and number of RIS elements has been presented here.