EUVIMEDEUVIMED entdecken
StudienprotokollOpen Access · CC BY 4.0

Multi-Agent Deep-Reinforcement-Learning-Driven Energy-Aware Geographic Routing Protocol for UAV Ad Hoc Networks

Yanan Zhang, Bo Ouyang, Haibo Huang, Kun Zou, Zhao Huang, Rui Hu

Eng · 14.11.2026 · 7 (11) · 646

DOI: 10.3390/eng7110646 ↗

Kurz gefasst

Studienprotokoll in der Zeitschrift Eng, online erschienen am 14.11.2026. Verfasst von Yanan Zhang und 5 weiteren Autor:innen. Die Arbeit nennt 33 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Shiyan Key Laboratory of Electromagnetic Induction and Energy-saving Technology, Wuhan Science and Technology Major Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Unmanned aerial vehicle ad hoc networks (UANETs) are a key enabler of air–ground integrated information networks, yet their deployment is hindered by dynamic topologies with relative speeds often exceeding 30 m/s, stringent onboard energy constraints, and harsh propagation environments. Existing geographic routing protocols reduce control overhead but base forwarding decisions on position alone, and reinforcement-learning-based variants adopt single-agent architectures in which nodes learn independently; neither the fairness of energy distribution among neighboring nodes nor dense hop-by-hop feedback is considered. This paper proposes a Multi-Agent Deep-Reinforcement-Learning-Driven Energy-Aware Geographic Routing Protocol for UAV Ad Hoc Networks (MA-EAGP). Each node acts as an independent agent within the centralized-training-with-decentralized-execution framework, and all nodes share a single Actor network. A neighborhood energy fairness (NEF) index based on Jain’s fairness index extends energy awareness from a single next hop to its forwarding region, while a comprehensive link quality (LQ) score fusing SINR, RSSI and ETX characterizes instantaneous link reliability. A dense distance-progress reward supplies immediate directional feedback, mitigating the sparse-reward problem and suppressing detours. Simulations across node densities and speeds show that, in most of the evaluated scenarios, MA-EAGP improves the packet delivery ratio and network lifetime, yielding lower end-to-end delay and fewer dead nodes than GPSR, QNGPSR and DDQN-MTGPSR.

Förderung

  • Shiyan Key Laboratory of Electromagnetic Induction and Energy-saving Technology
  • Wuhan Science and Technology Major Project

Zitieren

APA

Zhang, Y., Ouyang, B., Huang, H., Zou, K., Huang, Z., & Hu, R. (2026). Multi-Agent Deep-Reinforcement-Learning-Driven Energy-Aware Geographic Routing Protocol for UAV Ad Hoc Networks. Eng, 7(11), 646. https://doi.org/10.3390/eng7110646

Vancouver

Zhang Y, Ouyang B, Huang H, Zou K, Huang Z, Hu R. Multi-Agent Deep-Reinforcement-Learning-Driven Energy-Aware Geographic Routing Protocol for UAV Ad Hoc Networks. Eng. 2026;7(11):646. doi:10.3390/eng7110646

BibTeX
@article{Zhang2026,
  author = {Zhang, Yanan and Ouyang, Bo and Huang, Haibo and Zou, Kun and Huang, Zhao and Hu, Rui},
  title = {Multi-Agent Deep-Reinforcement-Learning-Driven Energy-Aware Geographic Routing Protocol for UAV Ad Hoc Networks},
  journal = {Eng},
  year = {2026},
  volume = {7},
  number = {11},
  pages = {646},
  doi = {10.3390/eng7110646}
}
Kostenlos starten →