Joseph Ozanne is a specialist in the simulation of racing yachts, as well as optimisation and performance analysis. He is currently participating in his 5th America's Cup campaign for Alinghi Red Bull Racing, the professional sailing team representing Switzerland. This comes after his participation in 4 previous editions of the well-knowned America's Cup, two of which were won by American teams in which his AI work was involved.
You describe yourself as a specialist in the simulation of racing yachts. What does this involve and what role does AI play in your field?
Competitive sailing, and particularly the America's Cup, is comparable to Formula 1 in terms of research and development. As such, we make extensive use of simulators for both boat design, sailor training and performance optimisation. Very advanced models of the yacht are produced, both in terms of CFD (Computational Fluid Dynamics) and structure/deformation, hydraulic systems, etc. In this context, AI is used as a toolbox to solve problems such as data set fitting (deep learning), reinforcement learning (sailing bots), performance optimisation or even to replace certain on-board sensors with computer vision.
What are the major difficulties you face in your work to analyse and optimise the performance of racing yachts using AI ? How do you overcome them?
AI is used in a very applied way, with very little time spent on pure research. As a result, the main issue is - as with all AI - the quality of the datasets. As our sailing boats are operating in relatively hostile environments (wind, waves, deformations), the data collected is very often insufficient for quality predictions. The solution lies in the intensive use of simulated data, both for physical models (aerodynamics and hydrodynamics) and for global performance models (complete boat), as well as in support of on-board measurement, to limit uncertainty.
What are other possible applications of AI in the field of sports competitions?
AI can be used really anywhere there is data. Some sports are more suited to performance optimisation than others, and not all sports are at the same level of maturity in terms of their data culture and acceptance. But I am convinced that any practice can make significant progress as long as you can apply AI in a performance-oriented way. The objective in a sports competition is as simple as winning! But sportsmen and women are rarely scientists, and there are many preconceived ideas about the relevance of these approaches in the world of top-level sport - the adoption of AI in sailing is new and has a long way to go - but it will become essential in the future once it can be proven that its application to performance enhancement can be significant.