Ph.D. Defense: Victorien Romain – December 16, 2025

Victorien Romain, industrial Ph.D. student in collaboration with Orange Labs, is planned to defend his Ph.D. thesis on December 16, 2025, at 9:30 a.m., in the Amphitheatre Gaston-Planté, Accès 35 – Level 1, 2 rue Conté, 75003 Paris, France.

Thesis Title: A Data-Centric Approach for Next Generation Mobile Networks Managements

Thesis Title (French): Une approche basée sur les données pour la gestion des réseaux mobiles de nouvelle génération

Abstract:

Fifth-generation (5G) mobile networks introduced a service-based, cloud-native architecture that significantly increased the flexibility of network operations. Network Functions, the entities in charge of executing the operations, interact through a large number of standardized interfaces, generating large volumes of data. The standardization body 3GPP has specificed the Network Data Analytics Function (NWDAF) to support data-driven management. However, it still sees limited adoption by both the academic community and mobile network operators. 

This thesis proposes Publish-Subscribe Service-Based Architecture for Next-Generation (PSBA-XG), a data-centric publish/subscribe architecture designed to simplify and accelerate network management by continuously collecting, normalizing, and distributing 5G core network data. The platform captures service-based (HTTP/2) signaling through proxy-based interception, extracts user-plane information via PFCP, organizes all information into unified Kafka topics, and exposes it to analytics and control applications through a data pipeline.

In this thesis a data-centric approach is considered for the network management. A data-centric publish/subscribe architecture is proposed to simplify and accelerate network management by continuously collecting, normalizing, and distributing 5G core network data. A Publish-Subscribe Service-Based Architecture for Next-Generation (PSBA-XG) platform has been designed to demonstrate the possibilities and the gain of this approach. The platform captures service-based (HTTP/2) signaling through proxy-based interception and extracts user-plane information using PFCP protocol messages. The collected information are organized into unified Kafka topics and exposed to analytics and control applications through a data pipeline.

Using this platform, we demonstrate how a data-centric architecture allows for real-time visibility, fine-grained monitoring, and automated network control. First, the platform is used to optimize control-plane procedures, showing that the parallelized execution and early data availability can significantly reduce signaling latency and end-to-end execution time, while remaining fully compliant with 3GPP specifications. Then, the platform is extended to user-plane management, allowing for congestion detection in a scenario where the standard would not be able to, and enforce QoS control through a custom PFCP client that dynamically adjusts UPF rules based on aggregated traffic indicators.

Experimental evaluations on an open-srouce 5G core testbed, using two different implementations of the 5G core, confirm that PSBA-XG provides operational benefits with minimal added latency overhead. The results show the potential of our approach for future analytics in the mobile core network, in particular for the yet undefined term of “AI-native” used to describe 6G.

Victorien’s publications: