Carrier aggregation (CA) is an important technology component of 4G and 5G radio networks and it enables radio network node to use more than one carrier to transmit/receive the data to/from the mobile terminals. As operators are adding newer frequency bands to their deployments and as almost all the mobile terminals support the CA capability, it is a feature of utmost importance for an operator to handle the growing traffic demands.

The support for new frequency bands at the mobile terminals comes with a limitation that the mobile terminals support only a limited set of frequency band combinations for carrier aggregation amongst all the supported frequency bands. For example, a mobile terminal that supports 20 different frequency bands can have 2^20 different band combination possibilities amongst which only a handful of band combination would be supported by the mobile terminal. At any given point in time, most often a mobile terminal is in the coverage of only a few of those frequency bands with varying coverage quality levels and the cells on each of those frequency bands are unequally loaded in terms of data traffic. Continually configuring a mobile terminal with the best possible combination of serving cell set in these frequency bands would enhance the user experience however this comes at a constant cost of processing and decision making at the network side as the cell load keeps varying with every new user admitted into the network. Thus, the combination of continuous movement of a mobile terminal, its supported frequency band combinations and the varying cell loads on those cells which are candidates to be in the cell set for this mobile terminal makes the serving cell set selection an extremely complex operation.

Thesis Description

This thesis aims to investigate the possibility to develop algorithms to perform the optimal serving cell set selection. The thesis also includes the usage of AI for solving the problem. The thesis involves usage and enhancement of a simulator to perform the studies. The thesis can develop in different directions according to the preferences and competences of the applicant, including:

  • Development of AI/ML models for serving cell set selection.
  • Development of rule-based algorithms for dynamic serving cell set selection.
  • Comparison of the performance and cost of AI based methods vs the rule-based algorithms.


We seek students with background in wireless communication and machine learning. The applicant needs to be a good programmer (Python, Matlab/Java) to manage the complexity of the task.


1-2 students, 30hp each

Preferred Starting Date

Spring 2024


Ericsson AB Mjärdevi, Linköping


5G, machine learning, optimization, python

Contact Persons

Samuel Axelsson                                                Pradeepa Ramachandra

+46 738 09 30 49                                               +46 725 07 40 75