Energy-Efficient and Stable Task Offloading and Resource Allocation in Space-Air-Ground Integrated Networks

Task scheduling in sixth-generation (6G) space-air-ground integrated networks (SAGIN) faces fundamental challenges due to limited terrestrial infrastructure, time-varying wireless links, and stochastic task arrivals. In such dynamic environments, task offloading and resource scheduling decisions are strongly coupled over time through queue evolution, which makes it difficult to simultaneously guarantee service stability and energy efficiency. To address this issue, we formulate a long-term stochastic optimization problem that minimizes the average system energy consumption while stabilizing task queues. Leveraging Lyapunov drift-plus-penalty optimization, we develop an online Four-Phase Computation Offloading and Resource Scheduling (FCORS) algorithm that decomposes the per-slot control into four tractable subproblems and operates without any offline training. Simulation results demonstrate that FCORS achieves performance parity with state-of-the-art Deep Reinforcement Learning (DRL) baselines. Specifically, under heavy traffic loads, it reduces energy consumption by approximately 51.7 (QOA) while suppressing queue backlogs by approximately 90.4 (GEA). Therefore, FCORS provides a robust and lowcomplexity solution for energy-efficient task scheduling in future 6G SAGIN-enabled MEC systems.