Dynamic Optimization of User and Cell Association in UAV-Assisted THz 6G Networks

The dynamic optimization of user and cell association in UAV-assisted THz 6G networks ensures high reliability, spectral efficiency, and real-time adaptability in high-mobility environments. However, beam misalignment, Doppler effects, and path delay distortions significantly degrade network performance. This paper introduces a novel machine learning-driven framework that integrates advanced optimization techniques for beam alignment, Doppler compensation, and user association. The proposed framework consist of three key algorithms: the Deep Reinforced Hybrid Angle Estimation Algorithm (DRHAEA), which employed deep reinforcement learning and transformer-based beamforming to dynamically adjust azimuth and elevation angles, mitigating beam squint effects and ensuring precise beam alignment; the Adaptive Neural Doppler Refinement Algorithm (ANDRA), which integrates Recurrent Neural Networks (RNN) and Multi-Stage Neural Networks (MSNN) to refine Doppler shift estimation and compensate for path delay distortions, thereby enhancing synchronization even in complex urban environments and the Neurodynamic Channel Tracker Algorithm (NCTA), which utilizes Deep Neural Networks (DNN), RNN, and Graph Neural Networks (GNN) for real-time channel tracking, efficient user clustering, and scalable connectivity. By mitigating beam squint, optimizing Doppler shift compensation, and improving user association, the proposed framework significantly enhanced spectral efficiency, reduced signal distortions, and improved UAV-ground communication reliability. Simulation results demonstrated that adaptive angle estimation and Doppler compensation effectively minimize beam misalignment and synchronization errors, achieving low-latency, high-throughput, and stable UAV communications. The framework achieved near-optimal spectral efficiency at SNR levels above 5 dB, outperforming conventional approaches. Comprehensive benchmarking against IEEE 802.15.3d and 6G performance metrics confirms its superior adaptability in dynamic THz network environments, making it a promising solution for next-generation UAV-assisted THz 6G networks.