A Machine Learning Approach for Predicting Pedaling Force Profile in Cycling

• DOI: 10.20944/preprints202408.0489.v1
• Publication Date: 2024-08-14T07:55:45Z
• Authors: Reza Ahmadi; Shahram Rasoulian; Samira Fazeli Vei...; Atousa Parsaei; Hamidreza Heidary; Walter Herzog; Amin Komeili
• Keywords: Cycling

Accurate measurement of pedaling kinetics and kinematics is vital for optimizing rehabilitation, exercise training, understanding musculoskeletal biomechanics. Pedal reaction force, the main external force in cycling, essential modeling closely correlates with lower limb muscle activity joint forces. However, sensor instrumentation like 3-axis pedal sensors costly requires extensive post-processing. Recent advancements machine learning (ML), particularly neural network (NN) models, provide promising solutions kinetic analyses. In this study, an NN model was developed to predict radial mediolateral forces, providing a low-cost solution study biomechanics stationary cycling ergometers. Fifteen healthy individuals performed 2-minute task at two different self-selected (58±5 rpm) higher (72±7 cadences. forces were recorded using system. The dataset included crank angle, cadence, power, participants' weight height. achieved inter-subject normalized root mean square error (nRMSE) 0.15±0.02 0.26±0.05 high respectively, 0.20±0.04 0.22±0.04 cadence. model's low computational time suits real-time predictions, matching accuracy previous ML algorithms estimating ground gait.