MODEL OF INTELLIGENT RESOURCE MANAGEMENT INTELECOMMUNICATIONS NETWORK

Abstract

The development of modern communication systems is characterized by a transition from traditional hardware architectures to flexible, software-defined environments. The implementation of 5G networks and the development of 6G concepts are fundamentally changing approaches to building mobile communication systems and telecommunication networks in general. The use of Software-Defined Networking (SDN) and Network Functions Virtualization (NFV) concepts leads to an exponential growth in traffic volumes and increased Quality of Service (QoS) requirements. Traditional static or heuristic resource management methods are unable to effectively adapt to sharp load fluctuations and the heterogeneity of network services in real time. In this regard, the application of Artificial Intelligence (AI) methods and Machine Learning (ML) algorithms to develop intelligent models for bandwidth allocation, routing, and queue management is a highly relevant task that allows for the automation of decision-making and the optimization of network performance.This paper develops a model for the intelligent resource management of a telecommunication network. The mathematical formulation of the optimization problem is presented as a utility maximization problem subject to bandwidth and delay constraints. The choice of the Deep Reinforcement Learning (DRL) algorithm for solving this problem in real time is justified, as classical mathematical methods are unable to cope with high traffic dynamics. A structural-functional scheme for integrating a DRL agent into the SDN architecture is proposed, providing flexibility in data plane management. The paper also presents the software architecture of the intelligent resource management model, including the specific software tools used (Mininet, Ryu, PyTorch). The software architecture of the developed system is based on the principles of modularity and loose coupling, which allows for the independent development, testing, and scaling of the intelligent component (DRL agent) and the network infrastructure. The proposed software solution consists of three main interacting modules: a network emulation module, an SDN controller module, and a machine learning module.