Modern 5G networks are like sprawling cities—full of traffic, constantly changing, and expected to run smoothly no matter what. To deliver ultra-fast speeds and low latency for things like autonomous cars,…
Modern 5G networks are like sprawling cities—full of traffic, constantly changing, and expected to run smoothly no matter what. To deliver ultra-fast speeds and low latency for things like autonomous cars, VR streaming, and IoT devices, networks rely on Radio Access Networks (RANs) and edge computing. But the problem is that RANs are now highly modular, which means that functions like the Distributed Unit (DU), Central Unit (CU), and User Plane Function (UPF) can be placed in different locations for flexibility. Edge servers have limited resources, and if these functions are not placed smartly, the network wastes power and slows down. Networks are also dynamic — new servers come online, old ones retire, and traffic patterns shift constantly.
A recent study by TITAN researchers at the University of Bristol introduces NetMind+, an innovative AI-driven framework designed to tackle this complexity. By combining Graph Convolutional Networks (GCN) with Deep Reinforcement Learning (DRL) and an incremental learning strategy, NetMind+ enables networks to adapt intelligently to changing conditions while minimising energy consumption.
NetMind+ composition and mechanism. Black and blue arrows correspond to static and dynamic DRL management, respectively.
NetMind+ uses maze-solving with reinforcement learning: imagining each network as a maze. The AI agent navigates this maze, deciding where to place functions and how to route traffic, aiming to minimise power use and meet strict latency requirements. As networks differ in size and shape, NetMind+ uses Graph Convolutional Networks (GCN) to encode each network into a standard format. This means one AI model can work across many different networks. It also utilises incremental learning. When the network changes (new servers, new links), NetMind+ updates itself layer by layer, cutting retraining time by 50%.
Key Numbers:
The implications of NetMind+ are significant: lower power consumption reduces operational costs and environmental impact, smarter placement ensures consistent performance for latency-sensitive applications, networks can adapt to evolving topologies without costly retraining, paving the way for self-optimising 5G and 6G systems.
As demand for high-speed, low-latency connectivity grows, solutions like NetMind+ will be essential for building resilient, efficient, and intelligent networks that power the next generation of digital services.
The research is a part of TITAN Lighthouse Project 1: Network Intelligence and End-to-End Integration. Lighthouse Project 1 (LP1) outcomes will help drive improvements in network performance, efficiency, and the ability to provide seamless and reliable services.
TITAN is part of the Federated Telecoms Hubs (FTH) alongside CHEDDAR, HASC, and JOINER. Together, we conduct world-leading R&D and create a new knowledge-sharing framework, roadmap technology, develop skills, and engage with standards while instilling a commercialisation-focused mindset across academia.
If you have any questions about the research discussed in this article, please contact titan-enquiries@eng.cam.ac.uk.
