Delay Optimization via Adaptive Computing Offloading Strategy for Time-Sensitive Tasks in Satellite Internet of Things

• Guisong Yang,
• Qiwen Xie,
• Panxing Huang,
• Xiangfei Li,
• Yechao Huang,
• Xingyu He,
• Yunhuai Liu,

Vol. 20, No. 1, January 31, 2026
10.3837/tiis.2026.01.009, Download Paper (Free):

• Adaptive offloading strategy
• Collaborative computing architecture
• Satellite internet of things
• Task offloading
• Time-sensitive task.

Abstract

In the Satellite Internet of Things (SIoTs), time-sensitive tasks generated by local Internet of Things (IoTs) devices in remote areas require offloading to other computing devices for processing because of their limited computing resources. In the traditional SIoT architecture, these tasks are sent to terrestrial cloud computing centers via satellite relay, leading to unacceptable transmission delay. With the growth of satellite technology, we propose a new SIoT multi-layer collaborative computing architecture, which utilizes collaboration among local IoT devices, ground edge servers, and a satellite cloud composed of satellites in different height. This architecture positions the cloud computing center closer to the edge. This architecture regards the satellite network as a cloud computing center, so as to minimize the end-to-end transmission delay for tasks. Given that the SIoT environment exhibits a time-varying nature and tasks are generated randomly, the traditional computing offloading strategy cannot update its strategy adaptively when facing the above environment, resulting in the increase of task processing delay. Against this backdrop, we describe the computing offloading problem within the proposed architecture as Constrained Markov Decision Process (CMDP) problem. Inspired by the advantages of deep reinforcement learning methods in dealing with dynamic environments, we design a time-sensitive task adaptive offloading strategy by integrating Deep Deterministic Policy Gradient (DDPG) to address the above CMDP problem, aimed at minimizing task processing delays. According to extensive simulation results, the adaptive computing offloading strategy herein can reduce the average task processing delay in the system. Compared with the baseline algorithms, this strategy can reduce the average delay by up to approximately 70%. Moreover, under varying task data sizes and different computing capabilities of ground edge servers, the proposed strategy consistently maintains the lowest delay.

Statistics

Cite this article