Evaluating Drift and Uncertainty in IMU-Based Bike and Rider Lean Angle Measurements During Downhill Mountain Biking

Accurate estimation of lean angles is crucial for analyzing athletes and their bikes in sports such as enduro and downhill mountain bike to assess performance and prevent falling. This study aimed to assess the drift of IMU-based lean angle estimations for both the rider and the bike and to assess the propagation of uncertainty from the IMU's gyroscope of bike lean angle measurements during real-world downhill and enduro mountain biking. Athletes were equipped with 2 IMUs, attached on the pelvis and sternum, to estimate the lean angle with OpenSim. The bike was equipped with an IMU on the frame to estimate the bike lean angle. Calibration movements were performed at the start and the end of the track to quantify drift by calculating the rms of the calculated angle around the two residual (non-leaning) axes during both calibration movements. Monte Carlo simulation was performed to assess the uncertainty. A difference in rms for the athlete lean angle between the two calibration movements was 4.2°, and values of the residual axes ranged from around -20° to 20°. For the bike, the difference in rms was found to be 3.0° and the values ranged from around -10° to 10° for the calibration movements at the end of the track. The Monte Carlo simulations showed the mean of the standard deviations to be 0.30°, which corresponds to an uncertainty of 0.60° when extended to 95%. Compared with an average difference between runs within athletes of 8.9° (±7.1°), this uncertainty is negligible. The results imply that angle estimation strategies in downhill and enduro mountain biking can be valuable and could be used in performance monitoring, and may also hold potential to be applicable in other fast-paced outdoor sports.