Project 6G-SMART

Abstract

The growing complexity of mobile networks has presented challenges in optimising the network performance. While 6G promises potential performance improvement, managing 6G operating to ensure optimal performance is achieved in all operating environment is challenging. This project deals with the challenges of 6G network self-organisation by utilising Open Radio Access Network (O-RAN) network architecture and machine learning (ML) techniques. In the literature, ML technique has been shown to be effective in optimising a specific aspect of a network. However, often multiple aspects of the network performance are required to be optimised simultaneously, and applying multiple ML models without proper management may result in conflicting decision making by each ML model causing the network to operate in undesirable state. To tackle this issue, the project aims to research solutions for coordinating multiple ML algorithms, develop a unified ML platform for O-RAN, and conduct real-world testing of these algorithms in an O-RAN environment. The project will first develop several individual ML algorithms to fulfil self-configuration, self-optimisation and self-healing. Then an integration of the ML models will be sought, and an ML orchestrator will be developed to manage and resolve conflicting decisions made by various ML models. Testing will be conducted in two countries focusing on both telecom and future smart factory environments.