British sailors trial uncertainty-aware AI to boost race tactics
A UK artificial intelligence company has begun trials of uncertainty-aware AI technology with British offshore sailors to enhance performance optimisation in live racing conditions.
digiLab, an AI company headquartered in the United Kingdom, has entered into a partnership with offshore sailing team Next Step Racing, comprising sailors Joss Creswell and Charlie Warhurst, to test machine learning systems during the 2025 Défi Paprec offshore race from Rouen to Roscoff. The trial will deploy what is described as uncertainty-aware machine learning on one-design Figaro Bénéteau 3 foiling yachts.
Technology integration
The focus of digiLab's system is to not only predict performance but also to provide an indication of the model's confidence and to flag any anomalies in real time. This aims to support crews in adjusting tactics as conditions change throughout the race, representing a departure from traditional maritime AI approaches which have largely centred on navigation safety rather than performance optimisation during live events.
Uncertainty-aware AI differs from standard predictive AI systems by quantifying how confident the system is in its predictions. This approach is considered critical in open-ocean racing, where unpredictable weather and sea conditions can introduce significant variability and risk.
Trial objectives
The testing, which takes place during a rigorous offshore racing event, is set to examine how this AI technology can assist human decision-making in high-variance and rapidly changing environments. Live data generated from the race will provide insight into the interaction between human judgement and artificial intelligence in extreme conditions.
"We're not trying to take the helm away from the sailors, we're giving them an extra layer of feedback - grounded in data, but aware of its limits. "What's exciting about this project is that it brings cutting-edge probabilistic AI into one of the most dynamic, high-variance environments imaginable. Offshore sailing is fast-moving and full of edge cases - making it the ideal setting to explore how decision-support systems perform under real pressure. It's a rare opportunity to move this kind of tech out of simulation and into the real world."
This comment was made by Professor Tim Dodwell, CEO and Founder of digiLab, who believes the value of such a trial lies in testing the technology outside simulated settings.
Performance data and route optimisation
digiLab's solution, called The Uncertainty Engine, employs probabilistic machine learning methods to model how a yacht is expected to behave under varying conditions, and to establish the associated confidence intervals. This supports key tactical decisions including sail trim and race strategy, aiming to augment rather than replace human navigators.
Joss Creswell, co-skipper of Next Step Racing with Charlie Warhurst, commented:
"It's brilliant to be partnering with digiLab. We're a young team and very excited to partner with an innovative and data driven company to push the limits of the sport and our understanding of weather routing. Offshore racing is very strategic, like chess, but it exists in a chaotic environment where uncertainty is everywhere. This is where we feel digiLab can make a difference and help us get to the standard of the more experienced sailors quicker.
"This system doesn't replace our instincts, it helps sharpen them during training. When you're out there, running on adrenaline and a few hours' sleep, having a system that flags anomalies and asks the right questions is a game-changer in terms of better understanding the weather routing software. It'll help us make smarter decisions, stay sharper for longer, and ultimately race better. And if that same technology can help commercial shipping become more sustainable? That's massive, because it means our racing and partnership is having an impact well outside the boundaries of the race course."
The collaborative project is also aimed at gathering data to inform potential future applications within commercial shipping. Even relatively small gains in efficiency, such as a 5 percent improvement in fuel use, could translate into significant reductions in CO2 emissions - estimated at 0.6 million tonnes annually in the UK, equivalent to the removal of approximately 300,000 petrol vehicles from the roads.
Future plans
Looking ahead, the team intends to expand the programme to include route optimisation under uncertainty, by integrating real-time weather data and employing advanced computer algorithms to identify and evaluate tactical options. There is also an ambition to trial decision-support interfaces in both high-performance sport and commercial marine operations.
The initiative between digiLab and Next Step Racing aims to demonstrate the utility of uncertainty-aware AI and its potential for adaptation to a range of maritime applications, with particular interest from the perspective of sustainability and operational efficiency in commercial shipping.