Building an ML-based Networking Data Analytics Function

Cloud-native 5G is an emerging topic that is enabled by the new architecture of 5G, cloud computing and virtualization. New service-based architecture of cellular networks allows core network components to be deployed as microservices , conformant to cloud-native principles. This approach paves the way for new research to efficiently automate, scale, and verify 5G core deployments.

Network Data Analytics Function(NWDAF)[1] is a 3GPP-defined network function in the 5G core that provides data analytics as a service to other network functions. It collects data from across the network, from the 5G Core and RAN, and possibly the transport network. Later, it processes the data acquired to generate analytics related to the network. These analytics can lead to several actions that can be made on the network, such as resource scaling, workload migration, anomaly detection and so on.  

The goal this thesis will be, to enhance a NWDAF, that is the result of a previous thesis, which is capable of collecting data from the 5G core network functions, with ML capabilities. Secondly, the student is expected to generate datasets using emulation tools and train NWDAF with them. Finally, the implemented solution will be evaluated. You will also have the opportunity to test your solution in a Private 5G O-RAN testbed and an operational campus network[2].

Prerequisites
- Interest in cellular networks
- Kubernetes and Open Telemetry experience is a plus
- ML model deployment experience is a plus

References: 
[1] https://www.ericsson.com/en/core-network/5g-core/network-data-analytics-function
[2] mediatum.ub.tum.de/doc/1841593/document.pdf

 

Contact 

Please send your application by email to Mehmet Mert Bese. Please include a grade report and your CV.