6G mobile networks will use machine learning algorithms that learn from data to make split-second decisions about wireless connections, rather than relying on rigid mathematical formulas. This matters because 6G will operate at higher frequencies than current networks, making signals more vulnerable to interference from walls, weather, or moving objects. Traditional communication models cannot adapt quickly enough to these unstable conditions. The researcher has already shown that custom machine learning training methods can improve wireless system reliability by 100 times compared to standard approaches. If successful, this work would allow 6G devices—from phones to autonomous vehicles—to maintain stable connections even in challenging environments. That reliability is essential when devices make decisions on their own, such as coordinating emergency services or managing industrial systems. The project will test these algorithms on real mobile phones, not just in simulations. The research also addresses a gap in how machine learning models report their own uncertainty. By understanding calibration error—how well a model’s confidence matches its actual performance—engineers can trust that a device knows when it might be wrong. Samsung and other industry partners will use the findings to shape future network equipment and standards.
View original technical description
Introduction: 6G, expected in the 2030s, aims to transform the communications landscape, merging the digital and physical worlds. 6G devices will evolve beyond mere communication tools, becoming integral to network infrastructure, driving data exchange, and decision-making. 6G will radically enhance connectivity, envisioning a more human-friendly, sustainable, and efficient society. However, it also presents substantial challenges. 6G will operate at higher frequencies to handle massive amounts of data from a surge of connected devices. These frequencies - more susceptible to interference from obstacles or atmospheric conditions - introduce unstable communication links. This is concerning as our reliance on clear communication grows. Since 6G devices will be making decisions on their own, they will need to communicate extremely reliably to enable critical, immersive, and omnipresent services. Crucially, as 6G integrates into our society, we should be able to trust the decisions that these intelligent devices make on our behalf. These challenges must be solved. Moving towards 6G requires transitioning from rigid mathematical models – found in traditional communication systems – to more agile, high-dimensional modelling techniques with real-time adaptation. Machine learning (ML) is key, offering flexibility by letting data, which describes the system and its performance, drive decision-making. By embracing ML, we can overcome past limitations, effectively adapting to 6G's dynamic nature for seamless communication and problem-solving. To harness ML's full potential in 6G, understanding how to exploit data to build well-optimised and reliable approaches is crucial. Current training strategies must evolve, with [9] emphasising integration of communication system domain knowledge into training procedures. Supporting this, our research in [10] created custom learning methods that incorporate reliability metrics in ML training to align model optimisation with real-world operational benchmarks. It was demonstrated that these tailored learning algorithms can bolster wireless system reliability by 100x compared to standard strategies. Aims and Objectives: This project aims to design ML algorithms tailored to enhance 6G wireless system reliability and apply these advancements on 6G devices, showcasing real-world progress. The key objectives include: Designing wireless system models tailored for 6G networks and amenable to machine intelligence (e.g., integrating intelligence into multi-user networks and existing cellular technologies). Developing novel ML solutions and custom training methods for each studied system to improve key communication reliability indicators. Understanding the calibration error (how the ML model's predicted and real-world performance align) associated with the developed ML solutions, ensuring accurate reflection of expected wireless system performance. Demonstrating real-world applicability of these ML solutions by deploying them on devices, e.g., mobile phones. Sharing and promoting research impact and contributions. Applications and benefits: Academia: Positions the host institution and academic partners (King’s College London, and Aalborg University Denmark) at the forefront of 6G research, gaining insights into technical/ethical considerations of ML. Industry: Samsung R&D Institute UK and Dhali Holdings Ltd benefit from research findings and innovations. Potential for job creation and economic growth. They receive a broader audience and business opportunities. Policymakers: Contributes to the UK’s Wireless Infrastructure Strategy and Industrial Strategy. Provides policymakers valuable insights for informed decision-making. Promotes project outcome adoption, furthering societal and economic growth. Public: More reliable wireless communication for essential services and emergencies. Focus on environmental sustainability, improving quality of life for various communities. Economy: Strengthens UK's digital communications framework. Positions UK as a global leader in 6G. Fosters job creation and socio-economic benefits in the UK.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know