کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

DPPO-Accelerated Two-Time-Scale Learning for Mobility-Aware Multi-Beam THz-NOMA Under Blockage

نویسندگان :

Khashayar Saremi Tabrizi¹ Fatemeh Alia² Alireza Mard Shoorijeh³ Bahareh Akhbari⁴

1- دانشگاه صنعتی خواجه نصیرالدین طوسی 2- دانشگاه صنعتی خواجه نصیرالدین طوسی 3- دانشگاه صنعتی خواجه نصیرالدین طوسی 4- دانشگاه صنعتی خواجه نصیرالدین طوسی

کلمات کلیدی :

Terahertz communications،power-domain NOMA،multi-beam small cells،blockage and mobility،two-time- scale control،deep reinforcement learning

چکیده :

Terahertz (THz) small cells empowered by power- domain NOMA (PD-NOMA) can meet 6G peak-rate demands, yet their performance is highly sensitive to mobility-driven angular dynamics, blockage, and beam misalignment. This paper studies mobility-aware resource management in a multi-beam THz-NOMA downlink, where each active beam serves a near–far user pair under SIC feasibility and outage constraints. We for- mulate a two-time-scale continuous control problem that updates beam selection and per-beam power budgets at a slower beam- management timescale, while adapting per-beam NOMA power splits every slot. To address the resulting high-dimensional, non- convex optimization under realistic THz dynamics, we develop a centralized Proximal Policy Optimization (PPO) agent and a distributed PPO (DPPO) variant to improve scalability and convergence. The environment integrates geometry-based block- age with hysteresis, molecular absorption, correlated shadowing, LoS/NLoS fading, AoD uncertainty, and codebook beamforming with inter-beam interference. Simulations under frozen scenarios and consistent evaluation protocols show that PPO/DPPO outper- form representative benchmarks in reliability-aware utility and spectral efficiency, demonstrating robust two-time-scale learning for THz-NOMA beam and power control.