Exploiting sensor data in professional road cycling: personalized data-driven approach for frequent fitness monitoring

We present a personalized approach for frequent fitness monitoring in road cycling solely relying on sensor data collected during bike rides and without the need maximal effort tests. use competition training of three world-class cyclists Team Jumbo–Visma to construct personalised heart rate models that relate exercise pedal power signal. Our model captures non-trivial dependency between exertion corresponding response rate, which we show can be effectively estimated by an exponential kernel. To daily are required day-to-day estimation, aggregate all sessions previous week apply sampling. On average, explained variance our is 0.86, demonstrate more than twice as large ignore temporal integration involved heart’s exercise. cyclist monitored tracking developments parameters models. In particular, monitor decay constant kernel involved, also analytically determine virtual aerobic anaerobic thresholds. findings threshold average agree with results believe this work important step forward performance optimization opening up avenues switching adaptive programs take into account current physiological state athlete.