Skip to main content

Project Context

Context

Networks

Networks have evolved significantly from traditional static infrastructures to more dynamic, intelligent, and adaptive systems.​

5G and Beyond-5G networks must:​

  • handle vast amounts of data.​
  • support a diverse range of applications. ​
  • ensure high reliability and low latency. ​

Goal: Achieve self-managing networks, where human intervention is minimized.​

NWDAF

Network Data Analytics Function​ are being developed with the goal of:

  • collecting and analyzing network data​
  • provide predictions for network optimization​

Has three main aspects:​

  • Data Collection​
  • Analytics Processing​
  • Analytics Exposure​

Goal: Automating the 5G network with machine learning and data analytics

MLOPs Pipeline

  • MLOps is an extension of DevOps, specifically adapted for machine learning workflows.​
  • End-to-end machine learning development process.​
  • Aims to unify the release cycle for machine learning.​
  • Enables the application of agile principles to machine learning projects.​

Problem

  • Increase in data consuming slows the network.​
  • Utilization spikes, can compromise network QoS.​
  • Technical problems (latency, packet loss) cause data transmission delays.​
  • Simply setting up more network and maintaining it is expensive.​

Goals

  • Implement a Data Pipeline for Network Intelligence.​
  • Develop and Integrate Machine Learning Models.​
  • Automate Network Optimization & Decision-Making.​
  • Implementation of a scalable and modular MLOps pipeline that works as a NWDAF when integrated with a 5G network.​
  • Ensure Compliance with 3GPP Standards. ​
  • Attempt Integration with existing 5G Core Network Components.​
  • Ensure seamless Communication Between Services within the pipeline.​
  • Evaluate System Performance under several Network Conditions.​
  • Provide a User-Friendly Deployment and Monitoring System.​

Expected Results

  • A modular and scalable MLOps pipeline that when integrated with a 5G network works as a base for a NWDAF.​
  • A NWDAF-like architecture that can be deployed in any 3GPP-compliant network. ​
  • Comprehensive validation of NWDAF-like architecture performance through real-world use cases in a 5G environment.