S4S - S.W.E.A.T. for Science

The question of how to produce effective and appropriate physical training plans is of great importance in many areas. For both professional and amateur athletes, getting the most out of every training session and progressing toward a long term fitness goal is of obvious interest. Even more important is the building of physical fitness for the less able-bodied. For cardiovascular patients and those in need of physical rehabilitation, the effectiveness of physical training is a quality of life issue of critical importance. Regardless of the application or goal, it is important to avoid injury or exhaustion due to training, and at the same time achieve the maximum long term performance improvement.

The Specific Workout, Efficient Assisted Training (S.W.E.A.T. for Science) project focuses on the modeling of training workout plans and the prediction of individual long-term performance under these plans. The aim is to automatically produce workout plans which deliver the largest long-term increase in performance, while avoiding exhaustion or unnecessary strain. These workout plans, combined with the intelligent control of electrically-assisted bicycles, facilitate appropriate training regardless of the length or ruggedness of the route, headwinds, or other factors. Based on individually tailored performance predictions and workout plans, the electric motor can provide additional assistance to prevent excessive stress, or lower the assistance to ensure adequate training. 

The S.W.E.A.T. for Science project specifically addresses the improvement of endurance training in running and cycling. Existing models from the sports science literature are examined and used as a starting point, and we apply own expertise in machine learning and signal processing to modify them and to develop models of our own. As we learn more about the physiological effects of training, we can integrate this knowledge into an adaptive e-bike controller to allow professional athletes, recovering patients, and everyone in between to get the most out of their training. Additionally, motivation aspects are relevant to develop a customized interface which can meet the individual needs of cyclists and adjusts itself adaptively to the user and support reaching their personal activity goals while avoiding overloading.