Convlutional Neural Network Based Hybrid Precoding For Cell Free Massive Mimo System

Abstract

Combining terahertz (THz) and cell-free massive Multiple-Input Multiple-Output (CFMM) communication technologies offers a viable solution to satisfy the high performance demands of next wireless networks. Significant advantages of this combination include reduced system complexity, increased energy efficiency, and improved spectrum efficiency. Notwithstanding these benefits, real-time signal processing and channel estimation in decentralized architectures are particularly difficult due to the dynamic and unpredictable nature of THz channels. Such rapid channel variations frequently make it difficult for conventional precoding and channel estimation approaches to react effectively. This Study presents a convolutional neural networks (CNN)-based precoding framework created especially for CFMM systems running at THz frequencies in order to overcome this problem. The suggested approach learns the spatial and temporal features of the wireless environment to dynamically build ideal precoding weights, utilizing the potent pattern recognition powers of convolutional neural networks. This adaptive architecture lowers computing overhead while greatly improving real-time responsiveness. The model's capacity to increase spectral and energy efficiency and provide scalable, reliable performance under various network situations is confirmed by simulation results. Simulation results show that, the CNN-based strategy outperforms traditional methods in terms of energy efficiency and system complexity at a signal-to-noise ratio (SNR) of 25 dB it exhibits improved scalability and resource optimization by achieving 17 bits per joule of energy efficiency, 580 bit/s/Hz of spectral efficiency and reducing system complexity to 1,100 FLOPS. These results demonstrate an energy efficiency improvement of approximately 41.67%, a spectral efficiency gain of 10%, and a 42.11% percent reduction in computational complexity compared to the conventional MMSE precoder. These outcomes highlight the potential of CNN-driven precoding as a transformative solution for THz-enabled CFMM architectures in next-generation wireless systems.