LEO Satellite-assisted Task Offloading for a Near Real-time Earth Observation Service

The next-generation regenerative payload-enabled low Earth orbit (LEO) satellites enable task offloading and delivering services to energy and computation-constrained devices in remote terrains. Recent studies on satellite-aided edge computing often focus on binary offloading scenarios, neglecting crucial system parameters such as service period, task deadline, and output size. To address these limitations, we propose a hierarchical computation framework for remote Earth observation-related services such as disaster prediction, 2D/3D scene observation, route finding, and rescue operations based on satellite images/videos. The proposed framework supports parallel and partial task offloading strategies, optimizing the communication and computation resources across the serving LEO satellite, adjacent LEO satellites, and cloud-aided gateway. Our objective is to minimize the worst-case task completion time, ensuring near real-time delivery of requested tasks. The formulated multi-time slot joint optimization problem is tackled via the proposed iterative algorithm based on successive convex approximation, demonstrating superior performance compared to baseline solutions.