Paris 2024: a bit of Inria at the Olympics!

When it comes to meeting the tight schedules of top-level athletes, researchers have their own Olympic motto: understanding, adaptation, solutions. Understanding is essential if you want to tackle a field problem and turn it into a research problem. In order to do so, scientists work directly with coaches and scientists from federations. “You first need to develop an in-depth understanding of their requirements and the methods they use. From a methodological point of view, tackling their real-life user problems in such a specialized anddemanding context is a really interesting work and contributes to our research”, explains Stéphane Huot, a contributor to the PerfAnalytics project and head of the Loki* project team at the Inria Lille - Nord Europe research centre.

Meetings are also held with federations in order to better characterize the problems being targeted. “These meetings give us the opportunity to put forward sensors that they can use to objectively measure performance. This is highly- pplied research, working closely with users”, says Lionel Reveret, head of the PerfAnalytics projet, and a researcher at the Inria Grenoble - Rhône-Alpes research centre.

The researchers from the PerfAnalytics project work with athletes during training in order to develop tools that will be operational within a very short timeframe. The aim is to gather as much information as possible on athletes’ movements during performance. For this, researchers employ techniques that are as transparent as possible so as not to interfere with athletes’ performance. For instance, they can't use markers that would get in the way during climbing. Among the possible applications, video is used to measure the quality of gymnasts’ posture, climbing grips or special pedals measure athletes’ efforts, etc.

The Kinovis experimental platform is also used for 3D analysis of body geometry. In climbing, for example, researchers use a 3D avatar of the athlete projected onto a video providing additional information, as in augmented reality. This provides a way of objectively measuring posture that is linked to a real-life situation.

The researchers adapt to athletes’ every need. Their tools are designed not only to help to understand performance, but also to devise strategies or to target areas for improvement.

For athletes, training is centred around breaking skills down into sub-skills. If you want to get quicker, running on its own isn’t enough: you also have to work on motor coordination, develop your muscles, etc. In order to improve overall performance, dedicated tools can be developed for each individual sub-skill. In the context of the Revea project, the MimeTIC project team has been using virtual reality to supplement training for athletes. In boxing, the tool can be used to work on anticipating attacks from your opponent. In “real life”, this sort of training is limited by the fact that athletes will actually sustain blows. With virtual reality, you can repeat exercises over and over again without the risk of injury. “It provides a range of stimuli that goes beyond reality. You can play on visual or audio perceptions or even take on an incredibly strong opponent, pushing the boundaries of conventional training”, adds Franck Multon, head of the MimeTic project team at the Inria Rennes - Bretagne Atlantique research centre.

Drawing on data from biomechanics, neuroscience and artificial intelligence, the Best-Tennis project, to which the Mouvement, sport, santé (M2S) laboratory at the University of Rennes 2 is a contributor, has the same objectives. It is focused on improving serves and returns in tennis, which are both critically important for success.

Athletes have shown a real appetite for these new tools, so much so that they are more likely to repeat virtual exercises like those conducted as part of the Revea project. Indeed, this is one of the major challenges for these collaborations: getting athletes to adopt these solutions. In the context of the PerfAnalytics project, the Loki project team is working on new modes of interaction and advanced interfaces for the semi-automatic annotation of performance videos. A machine learning algorithm will help human users to identify and mark each event making up a performance. “Ultimately, even if we develop good tools and an athlete wins a medal, we will never know how much of a contribution we’ve made”, says Stéphane Huot. However, the results from these projects will have a positive impact on other research with broader contexts and applications - which is a victory in itself!