Joint Online Optimization of Power Allocation and Task Scheduling for Data Offloading in LEO Satellite Networks

The rapid expansion of Low Earth Orbit (LEO) satellites inescapably leads to the explosive growth of space data in LEO Satellite Networks (LSNs). The stochastic nature of space data arrivals and the intrinsically time-varying satellite-ground links in LSNs pose significant challenges for offloading substantial volumes of space data from LSNs to the ground stations. To overcome these challenges, we systematically study the joint online optimization of power allocation and task scheduling for data offloading in LSNs. Firstly, we remove the constraint of mean rate queue stability from the formulated joint online optimization problem and leverage Lyapunov optimization to decouple it into a set of per-time-slot subproblems. Each subproblem is then divided into a task scheduling problem and a power allocation problem. Subsequently, we derive a closed-form optimal solution for the power allocation problem, and a multi-armed bandit-based quasi-optimal solution for the task scheduling problem. Furthermore, we extend the aforementioned solutions to address the original joint online optimization problem. Through theoretical analyses, we show that the proposed algorithms consistently attain a sublinear time-averaged regret. Extensive simulation results demonstrate that our proposed algorithms exhibit superior performance over other benchmarks.