Adaptive Residual Recurrent Neural Network with Heuristic Optimization for Spectral Energy Balancing in 6G Massive MIMO Systems

The development of 6G communication networks necessitates transformative advancements in massive MIMO systems to accommodate escalating data traffic and user demands. Different issues faced by the classical MIMO models are higher computational complexity, poor adaptability to dynamic environments, and suboptimal spectral-energy trade-offs. Classical algorithms often suffer from high computational complexity, limited adaptability to dynamic channel conditions, and suboptimal spectral-energy efficiency trade-offs. The primary objective of the research is to develop a hybrid precoding design using deep learning to optimize resource allocation and antenna selection in massive MIMO systems. Unlike classical telecommunication approaches, the implemented approach may achieve a superior trade-off between spectral and energy efficiency, setting a new benchmark for intelligent precoding strategies. Hence, to tackle several issues that takes place in the prior massive MIMO in 6G, a novel deep learning-based framework is designed by optimizing spectral and energy balancing in the 6G network for enhanced communication. In this research work, better spectral and energy balancing is performed using a novel technique, an Adaptive Residual Recurrent Neural Network (ARes-RNN), which is efficient to learn the structural information of the MIMO system along with the design of hybrid precoders. The applied Enhanced Dung Beetle Optimizer (EDBO) algorithm is used to optimize ARes-RNN parameters, enhancing the network’s learning ability and performance. Unlike the conventional models, the presented ARes-RNN model attained a spectral efficiency of approximately 79.4% for the SNR variation of 25 dB. The method shows improved energy and spectral efficiency balance, reduced computational complexity, and higher throughput. The performance of the 6G network in the massive MIMO is increased by the proposed deep learning with optimized parameters. The method achieves better spectral energy balance and is suitable for future wireless communication networks when compared to other classical approaches already existing in this domain.