Sites Inria

Version française

Sport Numericus

MDV - 25/06/2014

Enhancing sporting performance using analysis and simulation

The MimeTIC project-team at the Inria Rennes-Bretagne Atlantique research centre stands at the interface between two laboratories: the 'Movement, Sport and Health' laboratory in the science and technique of physical and sports activities, and the Institute for Research in Computer Science and Random Systems (Irisa) in IT. Its research activities focus on three main areas. MimeTIC Manager Franck Multon gives us the low-down.

Analysis

"Our first area of research is analysis . We study methods of improving movement. An athlete is always seeking to outdo themselves and achieve their best performance, running the risk of developing musculo-skeletal disorders. Some athletes thus find their careers brutally cut short following an injury. We are currently implementing techniques that allow us to measure movement and estimate inherent biomechanical constraints , in order to ascertain the risk for the person involved. Factors such as the level of effort exerted in each muscle, whether the repetition of a certain movement could prove dangerous and the rescheduling of training sessions to avoid overusing a joint are all considered. In partnership with the French Tennis Federation , the team analysed various types of tennis serve to find the least damaging for muscles and joints and to identify the most suitable serve to use in training.

To achieve this, we are developing what we term as musculo-skeletal models , which are numerical mathematics simulation models for the estimation of muscular effort using movement and external constraint measurements. This approach is particularly useful for muscles that we are unable to observe directly. In general, we measure the effort of certain muscles indirectly using sensors, i.e. by placing electromyographic electrodes on the skin and measuring the electrical activity from the muscle. However this only works for muscles close to the surface of the skin, making it impossible to measure deep-seated muscles. A needle is a possible solution but this hinders movement, making manoeuvre extremely difficult, and also inspires fear. Aside from the above, we have no means of exploration.

© MimeTIC- M2S - Inria

We have also worked with Bruno Mégret, a Tour de France doctor, on a medical condition that affects many top-flight cyclists: the sports hernia. This condition affects a deep-seated pelvic muscle, but there is no means of measuring its activity, particularly when cycling. Once again, the musculo-skeletal models allow an estimation of the muscular effort involved in different cycling exercises. The model incorporates measurements of some surface muscles in order to give the best possible estimation of the effort required by deep-seated muscles.

This work is the result of collaboration between specialists in biomechanics and IT. My STAPS* colleagues develop the models—the workings of joints, location of muscle attachments, etc.—while my Inria colleagues develop the computing engines enabling us to solve problems. The goal is to develop new numerical optimisation methods, numerically simulate these models, fine-tune the cost function used for numerical optimisation, put the models into action and simulate missing data, etc. In some ways, the biomechanical specialists set down the equation and the IT specialists offer effective solutions for solving it."

 

By developing biomechanical models and improving measurement and resolution methods, we are able to estimate muscular effort and help athletes improve their performance while avoiding injury and trauma. This is the analytical stage.

© MimeTIC - M2S - Inria

We are not developing new sensors but rather seeking to maximise existing system sensors using biomechanical models. With these models, we are able to overcome the shortcomings of sensors in certain complex situations and make more robust measurements. It also allows us to explore parameters not directly measurable with sensors, as described for the estimation of muscular tension using measurements of movement, i.e. pairing one sensor with another and adding a certain mathematical model will render more reliable findings.

In addition to preventing trauma, we are also working on enhancing performance . What type of tennis serve results in the best serve-volley sequence? The goal is to minimise the action time while maximising the speed of the ball. By analysing various serve strategies, we have developed methods of quantifying the impact of this action on performance. Once each strategy has been quantified in terms of performance, it is possible to give practical to indicators to trainers and athletes. In this instance, we analyse situations and extract information which allows us to say, "this is the most suitable action for this purpose".

Simulation

"The team's second area of research is the simulation of autonomous virtual humans . No measurements are required here; instead, our goal is to simulate the movements of a human being based on the limitations of their surroundings, to carry out a task. For instance, when off balance, how is balance restored, how are certain forces applied, will the necessary force be generated by pushing on an object, etc. As it stands, this research is still in its infancy. Simulating a movement that a human being would have performed naturally in the same situation remains a complex scientific challenge.  The eventual aim is to conduct prospective research on what could be the "optimal" action for a specific task in view of the limitations. We have not yet considered sporting 'tasks' as this is a very complex issue, particularly when faced with extreme limitations such as those found in sport.

The analysis consists of observing movement to deduce effective combinations. The simulation consists of enhancing that which exists through modification, adjustment or even the invention of new actions. We hope to reach this goal within a few years."

Analysis/simulation interaction

"Our third area of research combines analysis and simulation . On the one hand, we analyse a moving human and on the other hand, we simulate an autonomous virtual human. We then have them interact, in a dual for instance. Take the example of a goalkeeper in handball. One of the main difficulties is understanding which visual information is used by the goalkeeper to anticipate the trajectory of the thrower's ball, and deciding the action necessary to intercept it. We call this perception-action coupling . You have a thrower and a goalkeeper and you would like to examine how the goalkeeper reads the opponent to anticipate the trajectory of the ball. With real players, even the most experienced thrower will be unable to reproduce the exact same shot twice, if just one parameter is changed. However, if we seek to understand the influence of the thrower's wrist movement on the decision-making of the goalkeeper, we require two strictly identical situations where only the movement of the wrist changes at any one time. It is impossible in real life because we cannot control situations to that degree. There are so many variables in movement and the environment that it is very difficult to find a link between cause and effect.

We are now using virtual reality to take a different approach. This allows us to analyse the behaviour of the goalkeeper when faced with simulated throwers entirely under our control. In this way, we are able to repeat the exact same shot several times, with infinitesimal changes—of 5° or 10°—to the joint, and thus study its impact on the goalkeeper's decision-making. And statistically, we have seen that it is possible to observe the effect of these small changes, through repetition, on the goalkeeper's behaviour. The cause-and-effect relationship between an action and the goalkeeper's response can thus be analysed.

These experiments were conducted with the French Handball Federation , the Stade rennais football club, and Ulster Rugby in Belfast . The software developed through this collaboration is now routinely used in preparations for professional rugby players. Wearing a virtual reality headset, rugby players are now training themselves to detect tricks from the opponent; their reactions are later analysed and compared to those of other players in order to improve their anticipation abilities. The virtual human goes to move, we pause the simulation at any time and the player must predict to which side the model moves. The goal is to learn to correctly read the opponent's movement as quickly as possible. This is perception-action coupling.

Virtual reality is also a promising tool for learning motor skills . In virtual reality, it is possible to repeat the exact same situation several times, enhance the information environment or simplify the task (adding the physical limitations of our environment as the learner advances). However this poses new challenges. These days, for example, anybody can improve at the Nintendo Wii Tennis game, without actually improving their performance on a real tennis court. This raises the issue of how to ensure the transfer of skills acquired in virtual reality so that the motor skills developed are of equal value in real life.

© MimeTIC - M2S - Inria

In collaboration with the University of Brasov in Romania, we have studied how to learn the basketball free throw using immersive systems. One of the known problems with virtual reality is that distances are poorly perceived. This is particularly problematic when training for a basketball free throw, where the thrower is tasked with shooting a ball into a basket more than four metres away. This study aims to resolve this problem. In the study, we are attempting to ascertain the most appropriate form of visual feedback in order to eliminate distance perception issues in the free throw learning process.

*STAPS: science and technique of physical and sports activities (Sciences et techniques des activités physiques et sportives)

Keywords: Visual perception Sport numericus MimeTIC M2S Health Virtual reality

Top