2025/05/26 by Min Wu, Kefeng Guo, Xingwang Li +4
Computer Science · Engineering · #Advanced Memory and Neural Computing #Ferroelectric and Negative Capacitance Devices #Quantum-Dot Cellular Automata
paper · doi:10.1109/tcomm.2025.3573454
openalex publication_date 2025/05/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/29
Given the rapid growth of diverse communication demands, future large-scale satellite-aerial-terrestrial integrated networks (SATINs) need to simultaneously provide services to users while guaranteeing spectral efficiency, secrecy and energy efficiency. This paper addresses the problem of maximising secrecy energy efficiency (SEE) in SATINs, which can accurately describe the effective trade-off between security, spectral efficiency and transmit power. Particularly, we investigate a secure beamforming (BF) scheme in cognitive SATINs that employs rate-splitting multiple access (RSMA) and reconfigurable intelligent surface (RIS) in the presence of multiple eavesdroppers (Eves) in a UAV-aided secondary network (SN). To optimize the SEE for secondary vehicle users while satisfying the constraints of primary users (PUs), we utilize deep reinforcement learning (DRL) to address the coupling between different optimized parameters based on the improved long short-term memory proximal policy optimization (LSTM-PPO) algorithm. The main innovation of this paper is to design a sophisticated reward function, action space, and state space according to each constraint to speed up the convergence. In addition, simulation results show that the proposed DRL-based optimization scheme exhibits significant advantages in terms of SEE compared with benchmark schemes, validating the effectiveness of this work.