Multi-Agent Deep Reinforcement Learning for Energy Efficient RIS-assisted CoMP-NOMA Networks

This paper proposes a novel solution for optimizing energy efficiency (EE) and maximizing rates in reconfigurable intelligent surfaces (RIS)-assisted coordinated multi-point (CoMP) non-orthogonal multiple access (NOMA) networks through a multi-agent deep reinforcement learning (MADRL) system. The approach focuses on optimizing key architectural parameters: the phase shifts of the RIS, power allocations of the base stations (BSs), and the number of CoMP users. Our MADRL system dynamically adjusts these parameters to maximize energy utilization while meeting quality-of-service (QoS) requirements. Through comprehensive performance assessments and extensive simulations, we demonstrate that the MADRL-based optimization achieves near-optimal results that closely approximate the outcomes of an exhaustive search method. This paper demonstrates the efficiency and practicality of the MADRL approach in solving complex non-convex optimization problems in wireless networks where traditional methods may fall short.