Project

Virtualised Resource Management in O-RAN

Building a cohesive, adaptive network infrastructure that unlocks the full potential of O-RAN architecture for next-generation communications.

Initial approach

This project addresses the challenges arising from the virtualisation of both the Radio Access Network (RAN) and key core network components within the O-RAN ecosystem.

As these virtualised elements increasingly coexist on the same Commercial Off-The-Shelf (COTS) servers, new complexities emerge around resource management, latency optimisation, fault detection and dynamic configuration.

By addressing these challenges, we aim to create a cohesive and adaptive network infrastructure that can harness the full potential of O-RAN’s open and flexible architecture for next-generation communication systems.

To achieve this, we use advanced AI/ML-based solutions to:

  • Enable seamless integration.
  • Optimise performance.
  • Improve the reliability of these components.

Use case

Key Performance Indicator (KPI) prediction is a highly promising use case within the Open RAN (O-RAN) ecosystem, using AI-powered analytics to proactively monitor, predict and optimise network performance.

As disaggregated and virtualised RAN environments become increasingly complex, traditional reactive approaches to performance management are no longer sufficient. Predictive KPI analytics helps address this limitation, enabling operators to improve quality of service, reduce downtime and optimise resource utilisation.

KPI prediction in O-RAN presents several significant challenges, including:

  • The volume and variety of data generated by disaggregated, multi-vendor RAN components. Managing this data can be complex and requires robust systems for data collection, integration and processing.
  • The need for immediacy: Predictions must be generated quickly enough to enable real-time corrective action and minimise any impact on network performance.
  • The adaptability and generalisation of AI/ML models: Prediction models must maintain their accuracy across different network conditions, configurations and multi-vendor deployments.

Overcoming these challenges is essential to unlocking the full potential of predictive analytics within the O-RAN ecosystem and moving towards more intelligent and proactive network management.

lightbulb

Innovation focus

Our work on KPI prediction focuses on developing advanced xApps powered by state-of-the-art Long Short-Term Memory (LSTM) models to predict critical network performance metrics.

LSTM models are particularly well suited to analysing the time-series data generated by Open RAN (O-RAN) systems, thanks to their ability to capture temporal dependencies. Using these models, our xApps can predict key metrics such as latency, throughput, packet loss and resource consumption with a high degree of accuracy.

This predictive capability enables proactive network management, helping operators address potential issues before they affect performance.

The approach not only improves the efficiency and reliability of O-RAN systems, but also supports the deployment of intelligent, adaptive and dynamic networks capable of responding to the demands of next-generation services.

Let’s talk

We look at every challenge in its real-world context to determine what’s worth building — and the right way to build it.

Tell us about your project