Estimating 5G Cell Utilization by Passive Measurement: A Machine Learning Approach
Dominik Krah, Stefan Farthofer-Oster, Peter Dorfinger (2023): Estimating 5G Cell Utilization by Passive Measurement: A Machine Learning Approach In: 2023 International Symposium on Networks, Computers and Communications (ISNCC)
Utilization of a mobile communication cell’s physical resources is a crucial impact factor on the data rate and quality of service available to a network user. Information about the utilization is therefore of great interest to researchers and network providers. Measurement scenarios that should be carried out independent of network providers require a measurement method that does not depend on a connection to a base station. In this work, we propose a novel estimation algorithm and introduce a measurement tool based on this algorithm, which is able to estimate the utilization of a 5G cell’s physical resources without being connected to a base station. The proposed estimation algorithm is a two-step machine learning algorithm, based on Expectation-Maximization. An off-the-shelf software-defined radio is used for the implementation of the measurement tool, making it inexpensive and widely applicable. The measurement method is entirely passive and does not depend on any connection to the network, making it completely independent of network providers. Furthermore, we present validation measurements to demonstrate the practical applicability of the proposed method. During these validation measurements, the proposed algorithm was able to estimate the cell utilization with a mean absolute error of 0.12 over all measurements and utilization levels.
