Safeguarded AI for energy savings in Radio Access Networks
Safeguarded AI for energy savings in Radio Access Networks
What problem did this project address?
Mobile networks consume enormous amounts of energy, with the Radio Access Network (RAN) responsible for more than 70% of an operator’s power use. As data demand grows and network layers accumulate (2G–5G), energy costs and carbon impact continue to rise.
Even though RANs often operate with significant redundancy, deciding when parts of the network can safely switch off is extremely difficult. Traffic patterns vary by geography, time of day, and user behaviour, and switching off a single carrier affects neighbouring cells and user experience.
Today’s methods are conservative and inefficient, typically static, threshold-based rules that achieve minimal energy savings. Operators need a way to reduce energy consumption without compromising reliability or Quality of Experience (QoE).
Project Overview
Project Overview
What did we build?
We set the scene for the development of a safeguarded AI that revolutionises the way energy is saved in RANs and promotes the economic and environmental sustainability of future-generation mobile networks. The project prepared all the constituents required for the design, training, validation, and testing of AI-based solutions for radio element switch-on/off, and created realistic scenarios for the credible assessment of such solutions.
To do this, we built several core components:
Generation of synthetic network datasets that accurately reflect real traffic patterns without exposing sensitive operator data.
User handover models to understand how traffic shifts when a radio element is switched off.
Metrics for energy savings and QoE to evaluate whether an AI decision is both efficient and safe.
A curriculum of increasingly realistic test environments from single-carrier scenarios to large-scale, multi-site deployments involving thousands of elements.
All these components enable us to design, test, and validate trustworthy AI algorithms that can automate energy-saving decisions in real networks.
Why is this important?
This project tackled one of the telecom industry’s most urgent challenges: reducing energy consumption without compromising service quality.
If operators had access to safe, high-confidence AI for managing radio resources, they could:
Cut RAN energy use by up to 60%, far beyond what today’s basic policies achieve.
Reduce operational costs and carbon footprint at scale.
Deploy AI-powered energy controls with confidence, backed by quantitative safety guarantees.
Support global sustainability targets while maintaining seamless service for billions of users.
The outcome will set a precedent for how AI can be introduced confidently into safety-critical, high-impact infrastructure not only in telecom but also across energy, financial, and transportation systems.
About ARIA
ARIA — the UK Advanced Research + Invention Agency — funds high-risk, high-reward research aimed at solving nationally important challenges. Their Safeguarded AI programme brings together leading academic and industry teams to create AI technologies that are powerful, trustworthy, and safe to deploy in the real world.
As an ARIA Creator, Net AI contributed to one of the Safeguarded AI Programme’s flagship goals: demonstrating that AI can operate safely in a mission-critical domain like telecommunications, where both service reliability and energy sustainability are crucial.
