Utility Loss of Information-Optimal for Semantic Empowered RSMA in Satellite-Integrated Internet
Mengya Lu,Jianhao Huang,Tao Yang,Ye Wang,Jian Jiao,Qinyu Zhang
DOI: https://doi.org/10.1109/jiot.2024.3452317
IF: 10.6
2024-01-01
IEEE Internet of Things Journal
Abstract:Satellite-integrated Internet can provide pervasive intelligent services for ubiquitous terrestrial equipments (TEs) in the forthcoming sixth generation network. Consider that the most of existing multicast systems in satellite-integrated Internet are content-independent, which may result redundant data transmission in satellites with limited resources, we propose a semantic empowered rate splitting multiple access (RSMA) downlink system. First, we propose a semantic empowered metric, named utility loss of information (UoI), which can simultaneously capture freshness, mismatch of transceivers, and environment ingredient for the RSMA downlink system. Then, we design a joint content-and environment-aware sampling policy for discrete multi-state Markov sources to achieve minimum UoI, and provide rigorous proof to show the policy has a threshold structure and derive the closed-form transmission ratio. Further, we formulate a joint optimization of power allocation and rate control for the RSMA downlink system to minimize long-term average UoI with limited onboard resources, and utilize the Lyapunov optimization framework to transform the above problem to minimize the upper bound of corresponding drift-plus-penalty expression, and solve via a deep reinforcement learning-based algorithm. Simulation results validate that our scheme achieves the minimum long-term average UoI, under optimal tradeoff among timeliness, reliability, and environment-aware importance, and outperforms the state-of-the-art schemes.