JoWUA

Abstract

The importance of Massive Multiple-Input Multiple-Output (Massive MIMO) technology is an essential feature of today 5G and the future 6G wireless communication like it is capable of making the differences in spectral efficiency, network capacity, and the number of users it can serve significant. Based on joint optimization of energy efficiency (EE) and system throughput, also known as spectral efficiency (SE) is a problem that is difficult to address due to high-dimensional state space, dynamic channel conditions, inter-user interference, and hard power constraints. Conventional optimization models which include convex optimization and heuristic scheduling algorithms do not lend themselves well to the fast-paced conditions of wireless networks and a large number of antennas. This paper suggests a Deep Reinforcement Learning (DRL)-based optimization system that can enhance energy efficiency and throughput in Massive MIMO systems simultaneously. The resource allocation problem is called a Markov Decision Process (MDP) in which the agent learns the optimal policies to deal with beamforming, power distribution, antenna activation, and user scheduling by interacting with the network environment. The framework mainly uses the Proximal Policy Optimization (PPO) algorithm and is compared to the traditional methods of convex optimization and heuristic baselines. Large-scale simulations were carried out based on Rayleigh fading channel models, realistic base station-user topologies and antenna configurations up to 64 to 512 antennas. The results of the experiments prove that the proposed DRL framework plays a significant role in enhancing the performance of the network. As an example, a DRL model with 128 antennas had 8.5 bits/Joule energy efficiency and 8.2 bps/Hz spectral efficiency, compared to 6.7 bits/Joule and 6.1 bps/Hz with convex optimization. The difference in throughput between different levels of user density increased by 15-25 percent and the QoS violation rates were kept at below 1 percent and inference latency at less than 5 ms, which allowed real-time deployment. In general, the findings establish that DRL offers a scalable and adaptive optimization strategy that is efficient in balancing energy usage and throughput in large-scale wireless networks. The given framework is a step in the direction of the creation of self-optimizing, energy-conscious intelligent communication systems of the next-generation 6G networks.