Task Offloading for Space–Air–Ground Integrated Network Based on Iterative Double Auction Algorithm
Year: 2026
Type: article
Abstract: The rapid proliferation of artificial intelligence (AI) applications and fifth-generation (5G) communication technologies has resulted in an explosive growth of user devices and computation-intensive tasks, creating unprecedented demands for communication coverage and computational capacity. Conventional terrestrial networks, however, are often insufficient in disaster-stricken regions, remote mountainous areas, and maritime environments, where reliable connectivity and scalable computing resources are critically required. To address these limitations, the space–air–ground integrated network (SAGIN) has emerged as a promising architecture, combining terrestrial terminals, uncrewed aerial vehicles (UAVs) as flying base stations, and low Earth orbit (LEO) satellites as cloud-like servers to enable efficient task offloading and resilient service provisioning. Nevertheless, task offloading in SAGIN faces significant challenges due to the selfish behavior of participating nodes. Users and service providers may submit untruthful bids to maximize individual benefits, leading to inefficient resource allocation and degraded system performance. This paper presents a novel iterative double auction-based task offloading algorithm for space–air–ground integrated networks. The proposed mechanism establishes a double auction market between UAVs (as computational task buyers) and LEO satellites (as computing service sellers), where both participants submit bids for task offloading. A broker facilitates the iterative auction process, guiding participants to adjust their bidding and pricing strategies to maximize individual utility while promoting global social welfare. Moreover, a reinforcement learning-based scheme named as alternating Hedge algorithm is proposed. In alternating Hedge algorithm, participants can continuously improve their strategies through direct participation in the auction, without the need for involvement from the controller. Experimental results indicate the proposed method achieves quick convergence in the task offloading market and significantly enhances social welfare, offering an effective strategy for allocating resources in space–air–ground edge computing environments.
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7204437858