To content
Fakultät für Elektrotechnik und Informationstechnik
C. Bektas, S. Böcker, B. Sliwa, C. Wietfeld

Rapid Network Planning of Temporary Private 5G Networks with Unsupervised Machine Learning

-
in
  • Benjamin Sliwa
  • Dipl.-Ing. Stefan Böcker
  • Caner Bektas
  • Univ.-Prof. Dr.-Ing. Christian Wietfeld
  • Publications
  • CC5G.NRW
  • Plan & Play
  • SFB 876

In 2021 IEEE 94th Vehicular Technology Conference (VTC-Fall), Virtual Event, September 2021

Abstract:

Private 5G networks are considered key enablers for allowing industrial companies to deploy fully digitized production environments in an ad-hoc manner. Based on the foundation of exclusively assigned frequency spectrum resources, a guaranteed quality of service level can be achieved. However, strict requirements and regulations regarding antenna placement and interference with neighbors impose tough challenges on enterprises without expertise in communication network planning. For closing this gap, we propose an unsupervised learning-based network planning framework capable of rapidly and autonomously finding suitable solutions for antenna placement based on given environments and service quality, while satisfying regulatory restrictions. Results show that our proposed system reliably and rapidly calculates antenna positions and powers for realistic private 5G network scenarios. Temporary network deployments, e.g., Formula 1 tracks inside the given private 5G network, can be planned within minutes based on pre-calculated radio environmental maps.