A Novel Cross-Layer User Association and Traffic Offloading in TN-NTN CF-mMIMO Networks Using D3QN

Cell-free massive MIMO (CF-mMIMO) is a key enabler for 6G heterogeneous networks, offering high spectral efficiency and uniform user-experienced data rates. A fundamental challenge in CF-mMIMO lies in user association (UA), particularly as user equipment (UE) exhibits heterogeneous quality of service (QoS) demands that cannot be addressed by conventional UA schemes relying only on physical layer metrics. To address this challenge, we propose a cross-layer UA and traffic offloading scheme utilizing the integrated terrestrial and non-terrestrial (TN–NTN) systems. The scheme ensures QoS fairness between critical UEs (C-UEs) and non-critical UEs (NC-UEs) by dynamically associating them with terrestrial access points (TAPs) or offloading them to high-altitude platforms (HAPs) according to link quality and service requirements. The solution proceeds in three stages: initial access and link validation, service-aware pre-association to preserve resources for C-UEs while identifying NC-UEs suitable for offloading, and global optimization at the central processing unit (CPU) to finalize association and traffic distribution. The optimization problem is formulated as a mixed-integer non-linear program (MINLP) to maximize a network-wide utility function that accounts for HAP power cost and fairness. To solve the NP-hard problem efficiently, we employ a deep reinforcement learning (DRL) framework based on the dueling double deep Q-network (D3QN) with prioritized experience replay (PER) and target-network synchronization. Simulation results demonstrate that the proposed approach significantly improves fairness, load balancing, and QoS compliance compared with baseline UA schemes, providing a scalable and intelligent access strategy for dense 6G networks.