Computing-Aware Routing for LEO Satellite Networks: A Transmission and Computation Integration Approach

The increasing demands of remote sensing (RS) have put a strain on computation and transmission resources. Traditional ground-offloading techniques, which involve sending large amounts of raw data to the ground, suffer from poor satellite-to-ground link quality. Additionally, current satellite-offloading methods, which involve using low earth orbit (LEO) satellites within the visible range of RS satellites for processing, fail to fully utilize the computing capability of the network due to limited computation resources of visible LEO satellites, particularly in hotspot areas. To address these issues, this article proposes a novel computing-aware routing scheme for efficient offloading via LEO satellite networks. This scheme integrates transmission and computation processes and optimizes overall delay. By modeling the LEO satellite network as a snapshot-free dynamic network model, cross-time pathfinding with low memory consumption is enabled. Instead of shielding the dynamics, the proposed model converts the topology dynamics into the association dynamics between satellites and virtual nodes, which represents self-loops, special edge and node weights. The proposed computing-aware routing scheme processes tasks during routing instead of offloading raw data to ground stations, reducing overall delay and preventing network congestion. The routing problem is formulated as a combination of multiple dynamic single source shortest path problems and a genetic algorithm-based method is proposed to approximate solutions in a reasonable amount of time. Simulation results indicate that the proposed computing-aware routing scheme decreases overall delay by 78.31% compared to offloading raw data to the ground for processing when computing capability is 100 Giga floating-point operations per second, which is a commonly supported capability by most LEO satellites.