Maksymyuk T. Intelligent automated management of decentralized mobile communication systems

The thesis solves a scientific problem of the development of methods, models and means for the intelligent decentralized multi-operator infrastructure management in mobile networks, to increase its technical and economic efficiency for operators and users in the scenario of open spectrum and infrastructure markets. This work aims to resolve an uncertainty between the need of centralized management to enhance the technical efficiency of the mobile network, and the demand for decentralized open spectrum market with flexible business models for infrastructure deployment and sharing by operators to increase the economic efficiency of the mobile communication system. In particular, we have developed a concept of a decentralized mobile network that integrates the means of artificial intelligence and the blockchain technology to leverage the advantages of both centralized and decentralized business models in the mobile network, while eliminating their corresponding drawbacks and constraints. Such a solution allows to disrupt the mobile network market by enabling a trustable spectrum and infrastructure sharing among operators, underpinned by economic and legislative mechanisms. Thus, the key scientific contributions of the thesis are summarized as following. A new structural and functional model of decentralized mobile network is proposed to decouple the user plane, the infrastructure plane, the operator plane, the decentralization plane and the intelligent control plane to enable an automated coordination of the spectrum and infrastructure sharing among operators based on the smart contracts, the assets tokenization and the artificial intelligence. To enable the trading of a spectrum and an infrastructure among operators within the decentralized blockchain system, we propose a tokenization model of spectrum and infrastructure based on the non-fungible tokens, utility tokens and central bank digital currencies. Considering the demand for the ubiquitous connectivity and the upcoming paradigm of the micro-operators in 6G, we propose a new model of adaptive operator selection based on technical and economic criteria to enable seamless user connectivity across multiple operators with flexible SLA smart contracts, negotiated and executed in quasi-real-time. A particular attention in the thesis is given to the problem of decentralized coverage planning to allow operators extending the network coverage by placing new base stations in the most optimal way, taking into account the existing network infrastructure that has been built before. This challenge is solved by a modified version of the self-organizing maps with a combination of immutable and mutable neurons to exhibit the positions of existing and potential network infrastructure during the coverage optimization. An additional enhancement of the model has been done by replacing the conventional metric of Euclidean distance to the metric of SINR that allows optimizing the network coverage in a spatial domain, considering the impact of buildings in the urban environment on the coverage quality. To resolve the problem of spectrum management in a current semi-centralized scenario, we have developed a game theoretical model that allows big operators with a nationwide spectrum license to share it with micro-operators in a local scale based on the smart-contracts. Such a solution allows both improving the connectivity and experience of the end users as well as increasing profits of operators by sharing the spectrum based on evolutionary stable equilibrium strategies. In addition, the challenge of unlicensed spectrum sharing with interference mitigation is solved by the cooperative game model based on the trading of utility tokens among the operators in a blockchain. The proposed model converges to the Nash equilibrium within few iterations that ensures the fair utilization of unlicensed spectrum regardless of the number of operators. To further enhance the efficiency of spectrum sharing among operators considering their time-varying demands, we have developed the intelligent spectrum management model for mobile networks based on the recurrent neural networks, which allows to predict a traffic demand of the particular network service and allocate enough spectrum in advance. This allows operators to adjust their corresponding strategies to ensure the best tradeoff between technical end economic efficiency of the mobile network. Finally, we have proposed a method of intelligent data flows management in an optical transport network infrastructure. Proposed method implements a flexible data flows multiplexing and aggregation across wireless and optical domain to support an adaptive end-to-end quality of service provision for different network slices in the large scale decentralized mobile system.