Mobility-Aware Task Offloading and Resource Allocation in SAGIN-Based MEC for AIoT: A Deep Reinforcement Learning Approach
IEEE Transactions on Cognitive Communications and Networking · 2026 · 12 · 12977-12990
DOI: 10.1109/tccn.2026.3737898 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Cognitive Communications and Networking, 2026. Verfasst von Wenwu Zhu und 4 weiteren Autor:innen. Die Arbeit nennt 42 Literaturangaben. Gefördert von: National Natural Science Foundation of China Project, Joint Funds for Railway Fundamental Research of National Natural Science Foundation of China, National Key Laboratory of Ni&Co Associated Minerals Resources Development and Comprehensive Utilization.
Themen: Mobilisation
Abstract (Originalsprache)
The integration of Mobile Edge Computing (MEC) with Space-Air-Ground Integrated Networks (SAGIN) has significantly enhanced computing service capabilities for future Artificial Intelligence of Things (AIoT) applications. However, jointly optimizing task offloading and resource allocation remains a critical challenge due to the high mobility of network nodes and the significant heterogeneity of communication, computing, and storage resources. To address these challenges, this paper proposes a mobility-aware optimization framework tailored for SAGIN-MEC systems. Specifically, we first formulate a multi-layer heterogeneous system model to characterize the coupled dynamics of computation, communication, queue evolution, and node mobility. Subsequently, we propose TD3-LAORS, a Twin Delayed Deep Deterministic Policy Gradient (TD3)-based task offloading and resource scheduling algorithm that integrates Long Short-Term Memory (LSTM) networks and an attention mechanism. Specifically, the LSTM module extracts temporal dependencies from historical observations, while the attention mechanism selectively focuses on states critical to current decisions, thereby empowering the policy network to achieve superior offloading and resource allocation in highly dynamic SAGIN-MEC environments. Extensive simulation results demonstrate that TD3-LAORS significantly reduces both task processing latency and system energy consumption. Specifically, compared with state-of-the-art baselines, the proposed algorithm achieves up to a 52.4% reduction in system latency and a 39.1% decrease in energy consumption.
Förderung
- National Natural Science Foundation of China Project
- Joint Funds for Railway Fundamental Research of National Natural Science Foundation of China
- National Key Laboratory of Ni&Co Associated Minerals Resources Development and Comprehensive Utilization
- Key Project of Shenzhen City Special Fund for Fundamental Research
- Postgraduate Innovative Project of Central South University
- National Natural Science Foundation of Hunan Province
- High Performance Computing Center of Central South University
Zitieren
APA
Zhu, W., Deng, X., Li, W., Wan, S., & Min, G. (2026). Mobility-Aware Task Offloading and Resource Allocation in SAGIN-Based MEC for AIoT: A Deep Reinforcement Learning Approach. IEEE Transactions on Cognitive Communications and Networking, 12, 12977-12990. https://doi.org/10.1109/tccn.2026.3737898
Vancouver
Zhu W, Deng X, Li W, Wan S, Min G. Mobility-Aware Task Offloading and Resource Allocation in SAGIN-Based MEC for AIoT: A Deep Reinforcement Learning Approach. IEEE Transactions on Cognitive Communications and Networking. 2026;12:12977-12990. doi:10.1109/tccn.2026.3737898
BibTeX
@article{Zhu2026,
author = {Zhu, Wenwu and Deng, Xiaoheng and Li, Wang and Wan, Shaohua and Min, Geyong},
title = {Mobility-Aware Task Offloading and Resource Allocation in SAGIN-Based MEC for AIoT: A Deep Reinforcement Learning Approach},
journal = {IEEE Transactions on Cognitive Communications and Networking},
year = {2026},
volume = {12},
pages = {12977--12990},
doi = {10.1109/tccn.2026.3737898}
}
EUVIMED