Mobility-aware UAV framework employing hybrid optical and STAR-IRS for nonterrestrial networks

Abstract

This paper proposed a mobility-aware hybrid optical–radio frequency nonterrestrial network architecture employing unmanned aerial vehicles (UAVs) equipped with simultaneously transmitting and reflecting intelligent reconfigurable surfaces (STAR-IRS). The proposed framework integrates a multihop free-space optical (FSO) backhaul and an STAR-IRS–assisted RF access link to enhance link reliability, spectral efficiency, and robustness under dynamic mobility and atmospheric impairments. A novel UAV vibration-induced turbulence model is established to capture vibration-driven amplitude and phase distortions in hybrid channels. To jointly optimize UAV trajectories, beam alignment, STAR-IRS coefficients, and nonorthogonal multiple access power allocation, a two-phase successive convex approximation (SCA)–deep deterministic policy gradient (DDPG) framework is introduced, extended with a game-theoretic multi-UAV coordination mechanism based on Shapley value reward allocation and Nash bargaining for fairness. Simulation results demonstrate the proposed Game-DDPG approach significantly outperforms conventional SCA and DDPG baselines, achieving up to 10.8 % higher spectral efficiency. Under medium turbulence, the near-optimal deployment altitude is 220 m with high SNR of 30 dB, yielding BER < 10 ⁻¹⁸ and outage probability < 10 ⁻¹² under perfect channel state information. Under strong turbulence, Doppler, and vibration conditions, the system sustains BER < 10 ⁻⁸ and outage probability < 10 ⁻⁹ , demonstrating smooth degradation rather than abrupt failure. The system sustains the fairness index of 0.98, which implies the scalability and stability in multi-UAV scenarios.