Enhancing Multiuser Detection in Massive MIMO Systems with Self-Supervised GNN and Deep Reinforcement Learning
Year: 2026
Type: conference-paper
Abstract: This paper introduces a novel framework for enhancing Massive MIMO systems by integrating Graph Neural Networks (GNNs) and Deep Reinforcement Learning (DRL) to improve multiuser detection (MUD) and beamforming optimization. The proposed approach utilizes self-supervised GNNs for channel estimation, allowing for adaptive learning of interference patterns and refining the channel matrix. Additionally, DRL is applied to dynamically optimize the beamforming matrix, utilizing real-time feedback such as Signal-to-Interference-plus-Noise Ratio (SINR) and throughput. Through Monte Carlo simulations in a Rayleigh fading channel environment, the framework demonstrates significant performance improvements. The results show that the proposed model reduces Bit Error Rate (BER), enhances computational efficiency, and exhibits robustness to channel estimation errors compared to existing techniques. Furthermore, it handles dynamic user behavior and large-scale network configurations, offering a scalable solution for next-generation wireless systems.
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7169852832