Adaptive Ground User Pairing in Flying-STAR-RIS-Assisted AAV mmWave CF mMIMO Satellite-RSMA Networks

초록

This paper proposes an adaptive ground user pairing (AGUP) scheme for flying simultaneous transmitting and reflecting reconfigurable intelligent surface (STAR-RIS)-assisted autonomous aerial vehicle (AAV) millimeter-wave (mmWave) cell-free massive multiple-input multiple-output (mMIMO) satellite rate-splitting multiple access (RSMA) networks under imperfect channel state information (CSI). In the proposed scheme, two sets of ground users located in different areas are optimally paired to enhance spectral efficiency (SE). A max-min SE problem is formulated by jointly optimizing the transmit power, the position of the flying STAR-RISs, the binary user-pairing matrix, and the STAR-RIS phase-shift coefficients. Since the formulated problem is a mixed-integer non-convex program and thus difficult to solve directly, we employ an alternating optimization strategy that relaxes the original problem and decomposes it into three tractable subproblems-position, phase-shift design, and power allocation-solved iteratively using 2D Golden Section Search, coordinate descent, and an inner-approximation method. To enable real-time decision making, a deep reinforcement learning framework based on the deep deterministic policy gradient (DDPG) algorithm is developed, which jointly determines the optimal transmit power, STAR-RIS phase shifts, AAV position, and user pairing with substantially reduced computation time. Numerical results show that the proposed AGUP scheme yields 11.84% and 15.28% performance gains over random and non-ground user pairing schemes under RSMA, respectively, and achieves a 19.87% improvement over a NOMA-based scheme. Furthermore, the proposed DDPG approach demonstrates strong capability to predict near-optimal solutions while operating 99.95% faster than the alternating optimization-based AGUP scheme, making it highly suitable for real-time and dynamic satellite-terrestrial networks.

키워드