Hierarchical model of intelligent management of special purpose ground-air communication network

Abstract

In modern ground-air mobile networks, which include communication nodes with different technical characteristics, have a large dimension – thousands of users with requests for various services, one of the challenges is to ensure effective management. Such large-scale and dynamic nature of the operating environment creates a significant functional load on the management system, requiring the decomposition of management tasks. An effective solution is seen in the clear separation of goals into user and network goals, which will allow optimizing the network and meeting the needs of subscribers. The analysis of existing models of methods and algorithms for coordination, management and optimization in multi-level systems of Ad-Hoc type ground-air communication networks showed that they mostly have narrowly focused tasks of implementing individual target functions. In the article, for the first time, a hierarchical model of intelligent control of the ground-air communication network is proposed, which, from the point of view of the general structure, corresponds to the existing hierarchical model for the control of land mobile radio networks, but expanded by the description of the air network. In addition, the article shows the process of functional interaction between the node – node-metaagent – node-coordinator levels in the form of an algorithm. To optimize the process of interlevel interaction, it is proposed to apply machine learning algorithms with reinforcement. The article also mathematically describes the process of managing rewards and fines policies. It has been established that iterative methods provide greater flexibility and adaptability, but may require more time to find the optimal solution, however, in the case of early formation of a statistical sample of the functioning of individual control subsystems, it is possible to use non-iterative optimization methods at the initial stage of functioning, and reduce the time of the iterative process of retraining intelligent management systems of coordinator nodes and metaagent nodes