Author(s) :
R. Rejibha, Josma. J
Conference Name :
International Conference on Modern Trends in Engineering and Management (ICMTEM-25)
Abstract :
High-speed railway (HSR) communication systems demand reliable, low-latency, and high-throughput wireless connectivity in highly dynamic environments. However, rapid mobility, severe Doppler shifts, and dense user distributions introduce significant challenges in accurate channel estimation, inter-carrier interference (ICI) suppression, and scalable resource management. This work proposes an advanced framework that integrates lightweight deep learning with knowledge distillation for real-time channel estimation, and utilizes graph neural networks to exploit spatial correlations across users. To enhance ICI mitigation, edge reinforcement learning is deployed at remote radio units for adaptive interference suppression, supplemented by FFT-based pre-compensation techniques to counter extreme Doppler effects. Furthermore, federated transfer learning and meta-learning-based optimization are introduced to support decentralized model training and intelligent power allocation. Simulation results demonstrate improved normalized mean square error, enhanced spectral efficiency, and reduced computational overhead. The proposed system offers a scalable, adaptive, and efficient solution to meet the stringent requirements of next-generation high-mobility CF-MIMO networks.