EUVIMEDEUVIMED entdecken
Crossref

Reliable Session-Oriented Multi-Path Routing for LEO Satellite Networks: A Multi-Agent Learning Approach

Qi Guo, Yawen Tan, Tiago Koketsu Rodrigues, Nei Kato, Yohei Hasegawa, Masayuki Ariyoshi

IEEE Transactions on Networking · 2026 · 34 · 6606-6617

DOI: 10.1109/ton.2026.3722564 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Networking, 2026. Verfasst von Qi Guo und 5 weiteren Autor:innen. Die Arbeit nennt 47 Literaturangaben. Gefördert von: NEC x Tohoku Univ. Co-Creation Institute of Space-Integrated Network for Resilient Digital Transformation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Satellite networks are rapidly emerging as a key infrastructure for next-generation wireless communications, especially Low Earth Orbit (LEO) satellite networks, offering wide-area coverage, enabling low-latency communication, and improving support for emerging broadband and data-intensive services. However, their large-scale constellation, highly dynamic topology, intermittent inter-satellite links, and limited onboard resources pose challenges to reliable data delivery. In such environments, routing decisions should effectively cope with time-varying connectivity and constrained forwarding conditions. In this paper, we propose a multi-agent deep reinforcement learning-based routing framework for session-oriented data delivery in dynamic large-scale LEO satellite networks, aiming to improve delivery reliability under time-varying connectivity and constrained forwarding resources. This session-oriented routing solution is divided into two key cooperating processes: 1) a coarse-grained, learning-based multi-path routing decision algorithm; and 2) a fine-grained, efficient hybrid execution mechanism that ensures progress toward destination nodes in highly dynamic network environments. In this context, a centralized-training decentralized-execution multi-agent learning framework based on twin delayed deep deterministic policy gradient (TD3)-enhanced multi-agent deep deterministic policy gradient (MADDPG) algorithm is developed, where satellite agents independently make routing decisions according to locally observed network states. Additionally, a progress-aware mechanism with structured decision and fallback strategies is introduced to enhance the delivery reliability over time. Extensive experimental results, considering realistic orbital dynamics and generated on the MATLAB platform, demonstrate that the proposal significantly improves session delivery and reduces delivery latency compared with baseline routing methods.

Förderung

  • NEC x Tohoku Univ. Co-Creation Institute of Space-Integrated Network for Resilient Digital Transformation

Zitieren

APA

Guo, Q., Tan, Y., Rodrigues, T. K., Kato, N., Hasegawa, Y., & Ariyoshi, M. (2026). Reliable Session-Oriented Multi-Path Routing for LEO Satellite Networks: A Multi-Agent Learning Approach. IEEE Transactions on Networking, 34, 6606-6617. https://doi.org/10.1109/ton.2026.3722564

Vancouver

Guo Q, Tan Y, Rodrigues TK, Kato N, Hasegawa Y, Ariyoshi M. Reliable Session-Oriented Multi-Path Routing for LEO Satellite Networks: A Multi-Agent Learning Approach. IEEE Transactions on Networking. 2026;34:6606-6617. doi:10.1109/ton.2026.3722564

BibTeX
@article{Guo2026,
  author = {Guo, Qi and Tan, Yawen and Rodrigues, Tiago Koketsu and Kato, Nei and Hasegawa, Yohei and Ariyoshi, Masayuki},
  title = {Reliable Session-Oriented Multi-Path Routing for LEO Satellite Networks: A Multi-Agent Learning Approach},
  journal = {IEEE Transactions on Networking},
  year = {2026},
  volume = {34},
  pages = {6606--6617},
  doi = {10.1109/ton.2026.3722564}
}
Kostenlos starten →