EUVIMEDEUVIMED entdecken
Crossref

A SELF-LEARNING METHOD FOR RESOURCE MANAGEMENT IN TELECOMMUNICATION SYSTEMS USING REINFORCEMENT LEARNING

V. Gnatyuk, O. Lytvyniuk

Infocommunication technologies and electronic engineering · 2026-11 · 6 (2) · 18-27

DOI: 10.23939/ictee2026.02.018 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Infocommunication technologies and electronic engineering, 2026-10. Verfasst von V. Gnatyuk und 1 weiteren Autor:innen. Die Arbeit nennt 12 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a method for self-learning resource management in telecommunication systems based on Reinforcement Learning (RL). The proposed approach integrates predictive modeling of future network states, a multi-criteria reward function, and safety-aware action selection, enabling autonomous adaptation of resource allocation policies in real time. The architecture of the algorithm includes modules for telemetry collection, traffic and load forecasting, decision making, reward computation, and parameter updates within an Actor–Critic framework. A theoretical analysis of convergence properties and computational complexity was conducted of the algorithm’s convergence properties and computational complexity was conducted, demonstrating its suitability for real-time operational environments. Experimental evaluations were performed using simulations of telecommunication networks with dynamic network slices and virtualized network functions (NFV). The results indicate significant improvements compared to conventional and contemporary methods: reduction of average data transmission latency by up to 58%, packet loss to as low as 0.9%, a sixfold decrease in SLA violations, an increase in resource utilization efficiency up to 84%, and a reduction in energy consumption by approximately 25%. These findings confirm that the proposed method achieves a balanced optimization between network performance, resource efficiency, and SLA compliance, while adapting proactively to varying traffic patterns and network conditions. The integration of predictive modeling into the RL framework allows the system to anticipate network load changes and adjust policies proactively, contrasting with purely reactive control strategies. The methodology is fully compatible with Software Defined Networking (SDN) and NFV infrastructures, as well as next-generation 5G/6G networks and Internet of Things (IoT) environments, providing a scalable and adaptive solution for complex, distributed telecommunication systems. This work contributes both to the theoretical foundations of reinforcement learning applied to telecommunication resource management and to practical implementations for real-world network automation. It opens prospects for further research, including hierarchical multi-level network management, multi-agent distributed control, federated learning for privacy-preserving optimization, and low-latency, ultra-reliable applications. The presented method thus offers a robust framework for advancing autonomous network management and optimizing operational efficiency in modern telecommunication systems.

Zitieren

APA

Gnatyuk, V., & Lytvyniuk, O. (2026). A SELF-LEARNING METHOD FOR RESOURCE MANAGEMENT IN TELECOMMUNICATION SYSTEMS USING REINFORCEMENT LEARNING. Infocommunication technologies and electronic engineering, 6(2), 18-27. https://doi.org/10.23939/ictee2026.02.018

Vancouver

Gnatyuk V, Lytvyniuk O. A SELF-LEARNING METHOD FOR RESOURCE MANAGEMENT IN TELECOMMUNICATION SYSTEMS USING REINFORCEMENT LEARNING. Infocommunication technologies and electronic engineering. 2026;6(2):18-27. doi:10.23939/ictee2026.02.018

BibTeX
@article{Gnatyuk2026,
  author = {Gnatyuk, V. and Lytvyniuk, O.},
  title = {A SELF-LEARNING METHOD FOR RESOURCE MANAGEMENT IN TELECOMMUNICATION SYSTEMS USING REINFORCEMENT LEARNING},
  journal = {Infocommunication technologies and electronic engineering},
  year = {2026},
  volume = {6},
  number = {2},
  pages = {18--27},
  doi = {10.23939/ictee2026.02.018}
}
Kostenlos starten →