To content
Fakultät für Elektrotechnik und Informationstechnik

Dissertation on AI-based Network Slicing and Private Network Planning successfully defended

  • News
  • Type
  • News
© CNI, 2024
Caner Bektas - with the traditional doctoral hat designed by his colleagues - together with the examination committee

The CNI team congratulates Dr.-Ing. Caner Bektas on the successful defense of his dissertation entitled "Machine Learning-Enabled Dimensioning of Slicing-Based Private Mobile Communication Networks". Caner Bektas's dissertation addresses the challenging planning and dimensioning issues arising from the 5G innovations "campus networks" and "network slicing". As an innovative solution approach, he uses different machine learning methods, which on the one hand enable a predictive resource allocation for the realization of network slicing and on the other hand can support a demand-oriented network planning for the campus networks. As an additional framework constraint, he takes into account the resource limitations that will arise in future conceivable nomadic campus networks for applications in the event industry or agriculture.                 


With the introduction of 5G networks, 100 MHz spectrum was made available for so-called campus networks at the request of industrial users in addition to the frequencies for public networks. These campus networks allow private organizations, especially companies with high mobile networking requirements, to set up and operate their own 5G networks. This is accompanied by the challenge of efficiently planning this new type of local cellular radio network and complying with the strict regulatory requirements. 
Another innovation of 5G mobile technology is the ability to define virtually delimited areas, so-called "network slices", within a physical 5G network without these areas interfering with each other. In this way, application-specific performance profiles can be defined and guaranteed in both public and private networks. Here too, however, the challenging question arises as to how the slices within the network can be dimensioned to efficiently utilize the limited spectrum. Due to changing framework conditions, dynamic adaptation of the use of resources is desirable here. 

The results presented by Caner Bektas are not only of high practical relevance in the currently ongoing introduction phase of operational 5G campus networks and will also be made available to a broad user community via the 5G.NRW campus network planner. The work also makes important fundamental contributions to 6G research, for which the targeted use of artificial intelligence methods is a key component.