Joint Task Offloading and Power Control in Satellite Edge Computing Networks: A Stochastic Optimization Approach

The rapid growth of low earth orbit (LEO) satellite networks has driven the adoption of satellite edge computing (SEC) to perform onboard data processing. However, dynamic satellite movements cause fluctuating channel conditions, and limited onboard resources further challenge the balance between latency and energy efficiency, while unpredictable task arrivals introduce additional uncertainties. To overcome these issues, this paper proposes a comprehensive long-term stochastic optimization framework that simultaneously optimizes task offloading decisions and power allocation. By employing a Lyapunov optimization method combined with a virtual queue mechanism, the proposed framework transforms long-term latency constraints into manageable per-slot optimization goals, facilitating online decision-making even under uncertain task arrivals. An efficient algorithm based on the penalty dual decomposition (PDD) framework, integrated with quadratic transformations, and sequential convex approximations (SCA), is developed to solve the resulting non-convex optimization problem. Simulation results validate that the proposed approach significantly reduces energy consumption, consistently maintains latency within specified limits, and enhances overall system stability and service continuity.