Ayuda
Ir al contenido

Dialnet


Next-generation self-organizing networks through a machine learning approach

  • Autores: David Palacios Campos
  • Directores de la Tesis: Raquel Barco Moreno (dir. tes.), Isabel de la Bandera Cascales (codir. tes.)
  • Lectura: En la Universidad de Málaga ( España ) en 2018
  • Idioma: español
  • Tribunal Calificador de la Tesis: Carlos Camacho Peñalosa (presid.), Pablo Muñoz Luengo (secret.), Francisco García García (voc.)
  • Programa de doctorado: Programa de Doctorado en Ingeniería de Telecomunicación por la Universidad de Málaga
  • Materias:
  • Enlaces
    • Tesis en acceso abierto en: RIUMA
  • Resumen
    • At the present time, for most of the population, mobile phones have become the instruments through which to interact with the surrounding world. From its original service of voice transmission, cellular communications have evolved to provide a variety of services that could be hardly imagined just four decades ago. From that starting point, cellular networks were first enhanced to support data transmission, which opened the door to services like videocalls or web surfing. Later on, successive improvements were made so as to reach resource- and energy-efficient networks, while enhancing the users' perceived quality of experience (QoE) by means of an increasingly higher performance. In order to consequently reduce the management costs in such scenario, some further improvements needed to be made. This led to the concept of self-organizing networks (SONs). That is, the automation of the management tasks of a cellular network to reduce the operational and capital expenditure (OPEX and CAPEX, respectively).

      SON tasks are divided in three categories: self-configuration, self-optimization and self-healing. Self-configuration aims at automating the actions required when a network is to be deployed, like the initial configuration parameter setting. Self-optimization tasks pursue maximizing the efficiency of the networks in a time-varying environment, which takes shape as a variety of mechanisms addressing mobility, accessibility and integrity issues. Finally, the targets of self-healing are identifying and repairing possible failures that may arise while the network is operated.

      Thus, one of the main tasks of self-healing is determining the cause of a failure, which is called root cause analysis (RCA). Tools for RCA are automatic systems which, in the shape of classification systems, aim at determining a class (or network state) regarding a set of features (or key performance indicators, KPIs). Although different mechanisms have been proposed until now as tools for RCA, there is a long way to develop accurate systems that can deal with the large amount of performance information that is normally collected in a cellular network.

      Together with self-healing, self-optimization appears as the SON function group that attracts the most attention from industry and academia. This is mainly due to the optimization opportunities that novel functionalities from the upcoming networks bring. In particular, the management of multi-link connections, and within these, multi-connectivity (MC), occupies an eminent place in the next generation of mobile communications: the Fifth Generation New Radio (5G NR). However, given its novelty, multi-link communications currently lack of efficient management mechanisms, which will be one of the research hot topics in the coming years.

      The objective of this thesis is the improvement of SON functions through the development and use of machine learning (ML) tools for the network management. In particular, its target is twofold. On the hand, self-healing is addressed through the proposal of a novel tool for RCA, which takes the shape of a combination of multiple RCA baseline systems to develop an enhanced ensemble-based system. In order to further enhance the RCA accuracy while lowering both the CAPEX and OPEX, ML techniques for dimensionality reduction are proposed and assessed in combination with RCA tools. On the other hand, multi-link functionalities within self-optimization are studied, and techniques for automatic link management are proposed. In the field of enhanced mobile broadband (eMBB) communications, a component carrier manager implementing network operators' policies is proposed, whereas in the field of low-latency vehicular communications, a mechanism for multi-path traffic steering is proposed.

      Many of the methods proposed in this thesis have been assessed using data from live cellular networks, which has allowed them to demonstrate both their validity in realistic environments and their ability to be deployed in current and forthcoming cellular networks.


Fundación Dialnet

Dialnet Plus

  • Más información sobre Dialnet Plus

Opciones de compartir

Opciones de entorno