Abstract

Peak vertical ground reaction force is a common marker of running load, but it is normally measured on force plates that runners do not have outside the laboratory. This study asked whether peak force in runners excluded from model development can be estimated from marker-based kinematics alone, and whether a physical model of running adds to machine learning. Previously collected treadmill trials of 39 runners at 2.5, 3.5 and 4.5 m/s from a public biomechanics dataset gave 101 participant-speed trials. Outcome measures were derived exclusively from force signals, and predictors were derived exclusively from marker trajectories, with contact and flight times detected kinematically. Linear, kernel and tree-based learners were compared with a calibrated spring-mass estimate and with a hybrid in which machine learning corrected the spring-mass residual, using leave-one-runner-out validation with all tuning nested inside the training runners. Kinematic contact time agreed closely with force-based contact time (r = 0.96). The hybrid estimated peak force with a mean absolute error of 0.090 body weights and R² of 0.70, against 0.132 body weights for a speed-only baseline. The calibrated spring-mass model alone reached 0.090 body weights and was as accurate as the best purely data-driven model. When a whole running speed was withheld, the hybrid and ridge regression transferred better than either the spring-mass model or gradient boosting. Loading rate was estimated less well (R² 0.50). A spring-mass model with two fitted calibration coefficients therefore captures most of the information in kinematic measurements about peak running force, and machine learning should be judged against it.

Keywords

Running Biomechanics, Ground Reaction Force, Spring-Mass Model, Machine Learning, Leave-One-Subject-Out Validation, Explainable Modelling,

References

  1. Anderson, L.M., Martin, J.F., Barton, C.J., Bonanno, D.R. (2022). What is the effect of changing running step rate on injury, performance and biomechanics? A systematic review and meta-analysis. Sports Medicine - Open, 8(1), 112.
  2. Baker, L.M., Weigend, F.C., Swaminathan, K., Kim, D., Chin, A., Lieberman, D.E., Walsh, C.J. (2025). Estimating braking and propulsion forces during overground running in and out of the lab. PLOS ONE, 20(9), e0330042.
  3. Bernett, J., Blumenthal, D.B., Grimm, D.G., Haselbeck, F., Joeres, R., Kalinina O.V., List, M. (2024). Guiding questions to avoid data leakage in biological machine learning applications. Nature Methods, 21(8), 1444–1453.
  4. Binnekamp, M., Scheltinga, B.L., Wouda, F.J., Reenalda, J. (2026). No extra sensors needed: estimating vertical ground reaction force using IMUs at the smartwatch and heartrate strap location. Sports Biomechanics, 1–17.
  5. Burns, G.T., Tam, N., Langerak, N.G., Zernicke, R.F., Lamberts, R.P. (2024). Spring-mass characteristics in runners before and after a 56-km road ultramarathon. Journal of Applied Biomechanics, 40(6), 492–500.
  6. Carter, J., Chen, X., Cazzola, D., Trewartha, G., Preatoni, E. (2024). Consumer-priced wearable sensors combined with deep learning can be used to accurately predict ground reaction forces during various treadmill running conditions. PeerJ, 12, e17896.
  7. Chen, T., Xu, D., Zhou, Z., Zhou, H., Shao, S., Gu, Y. (2025). Prediction of vertical ground reaction forces under different running speeds: integration of wearable IMU with CNN-xLSTM. Sensors, 25(4), 1249.
  8. Donahue, S.R., Hahn, M.E. (2023a). Estimation of gait events and kinetic waveforms with wearable sensors and machine learning when running in an unconstrained environment. Scientific Reports, 13(1), 2339.
  9. Donahue, S.R., Hahn, M.E. (2023b). Estimation of ground reaction force waveforms during fixed pace running outside the laboratory. Frontiers in Sports and Active Living, 5, 974186.
  10. Evans, R.J., Moffit, T.J., Mitchell, P.K., Pamukoff, D.N. (2023). Injury and performance related biomechanical differences between recreational and collegiate runners. Frontiers in Sports and Active Living, 5, 1268292.
  11. Fukuchi, R.K., Fukuchi, C.A., Duarte, M. (2017). A public dataset of running biomechanics and the effects of running speed on lower extremity kinematics and kinetics. PeerJ, 5, e3298.
  12. Ghaffarzadeh, P., Chakraborty, D., Aslansefat, K., Dostan, A., Papadopoulos, Y. (2026). A multi-modal dataset for ground reaction force estimation using consumer wearable sensors. Scientific Data, 13(1), 855.
  13. Girard, O., Alsenoy, K.V., Li, S.N., Ryu, J.H., Peeling, P. (2022). Constant low-to-moderate mechanical asymmetries during a treadmill graded exercise test. European Journal of Sport Science, 22(4), 530–538.
  14. Gruber, A.H. (2023). The ‘impacts cause injury’ hypothesis: Running in circles or making new strides? Journal of Biomechanics, 156, 111694.
  15. Hata, K., Noro, H., Takeshita, T., Yamazaki, Y., Yanagiya, T. (2022). Leg stiffness during running in highly cushioned shoes with a carbon-fiber plate and traditional shoes. Gait & Posture, 95, 9–14.
  16. Honert, E.C., Hoitz, F., Blades, S., Nigg, S.R., Nigg, B.M. (2022). Estimating running ground reaction forces from plantar pressure during graded running. Sensors, 22(9), 3338.
  17. Johnson, C.D., Davis, I.S. (2022). What differentiates rearfoot strike runners with low and high vertical load rates? Gait & Posture, 96, 149–153.
  18. Kapoor, S., Narayanan, A. (2023). Leakage and the reproducibility crisis in machine-learning-based science. Patterns, 4(9), 100804.
  19. Koffman, L., Zhang, Y., Harezlak, J., Crainiceanu, C., Leroux, A. (2023). Fingerprinting walking using wrist-worn accelerometers. Gait & Posture, 103, 92–98.
  20. Leacox, A., Fashingbauer, L., Ferguson, T., Zajakowski, A., Baum, B., Reinking, M. (2025). The effect of running speed on cadence and running kinetics. International Journal of Sports Physical Therapy, 20(7), 957–963.
  21. Lemire, M., Falbriard, M., Aminian, K., Pavlik, E., Millet, G.P., Meyer, F. (2022). Correspondence between values of vertical loading rate and oxygen consumption during inclined running. Sports Medicine - Open, 8(1), 114.
  22. Li, S., Pan, J., Wang, D., Yuan, S., Yang, J., Hao, W. (2025). Predicting running vertical ground reaction forces using neural network models based on an IMU sensor. Sensors, 25(13), 3870.
  23. Ma, S., Zhang, J., Shi, C., Di, P., Robertson, I.D., Zhang, Z.Q. (2024). Physics-informed deep learning for muscle force prediction with unlabeled sEMG signals. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 32, 1246–1256.
  24. Malisoux, L., Gette, P., Delattre, N., Urhausen, A., Theisen, D. (2022). Spatiotemporal and ground-reaction force characteristics as risk factors for running-related injury: A secondary analysis of a randomized trial including 800+ recreational runners. The American Journal of Sports Medicine, 50(2), 537–544.
  25. Malisoux, L., Napier, C., Gette, P., Delattre, N., Theisen, D. (2023). Reference values and determinants of spatiotemporal and kinetic variables in recreational runners. Orthopaedic Journal of Sports Medicine, 11(10), 23259671231204629.
  26. Mayo-Nieto, C., Cartón-Llorente, A., Rafales-Perucha, A., Cardiel-Sánchez, S., Jaén-Carrillo, D., Roche-Seruendo, L.E. (2026). The biomechanical signature of elite distance running: A matched Pairs analysis across multiple speeds. Scandinavian Journal of Medicine & Science in Sports, 36(11), e70385.
  27. Meyer, F., Falbriard, M., Aminian, K., Millet, G.P. (2023). Vertical and leg stiffness modeling during running: Effect of speed and incline. International Journal of Sports Medicine, 44(9), 673–679.
  28. Miqueleiz, U., Aguado-Jimenez, R., Lecumberri, P., Garcia-Tabar, I., Gorostiaga, E.M. (2025). Reliability of inertial measurement unit-based spatiotemporal and kinetic variables in endurance runners during treadmill running. Journal of Human Kinetics, 97, 65–76.
  29. Mtibaa, K., Zarrouk, N., Ryu, J.H., Racinais, S., Girard, O. (2023). Mechanical asymmetries remain low-to-moderate during 30 min of self-paced treadmill running. Frontiers in Physiology, 14, 1289172.
  30. Nijs, A., Roerdink, M., Beek, P.J. (2023). Exploring running styles in the field through cadence and duty factor modulation. PLoS One, 18(12), e0295423.
  31. Nitschke, M., Dorschky, E., Leyendecker, S., Eskofier, B.M., Koelewijn, A.D. (2024). Estimating 3D kinematics and kinetics from virtual inertial sensor data through musculoskeletal movement simulations. Frontiers in Bioengineering and Biotechnology, 12, 1285845.
  32. Patoz, A., Gindre, C., Lussiana, T., Breine, B. (2026). The heel-strike fallacy and cadence beliefs in distance running: A narrative review. Sports Medicine - Open, 12(1), 149.
  33. Patoz, A., Lussiana, T., Breine, B., Gindre, C., Malatesta, D. (2022). Duty factor and foot-strike pattern do not represent similar running pattern at the individual level. Scientific Reports, 12(1), 13061.
  34. Patoz, A., Lussiana, T., Breine, B., Gindre, C., Malatesta, D. (2025). Comparison of different machine learning models to enhance sacral acceleration-based estimations of running stride temporal variables and peak vertical ground reaction force. Sports Biomechanics, 24(4), 825–841.
  35. Pitts, M.N., Ebers, M.R., Agresta, C.E., Steele, K.M. (2025). Evaluating sparse inertial measurement unit configurations for inferring treadmill running motion. Sensors, 25(7), 2105.
  36. Rasmussen, J., Skejø, S., Waagepetersen, R.P. (2023). Predicting tissue loads in running from inertial measurement units. Sensors, 23(24), 9836.
  37. Raymond, S.J., Cecchi, N.J., Alizadeh, H.V., Callan, A.A., Rice, E., Liu, Y., Zhou, Z., Zeineh, M. and Camarillo, D.B. (2022). Physics-informed machine learning improves detection of head impacts. Annals of Biomedical Engineering, 50(11), 1534–1545.
  38. Rosenblatt, M., Tejavibulya, L., Jiang, R., Noble, S., Scheinost, D. (2024). Data leakage inflates prediction performance in connectome-based machine learning models. Nature Communications, 15(1), 1829.
  39. Song, B., Paolieri, M., Stewart, H.E., Golubchik, L., McNitt-Gray, J. L., Misra, V., Shah, D. (2025). Estimating ground reaction forces from inertial sensors. IEEE Transactions on Biomedical Engineering, 72(2), 595–608.
  40. Turner, J.A., Chaaban, C.R., Padua, D.A. (2024). Validation of OpenCap: A low-cost markerless motion capture system for lower-extremity kinematics during return-to-sport tasks. Journal of Biomechanics, 171, 112200.
  41. Van den Berghe, P., Warlop, L., Derie, R., Leman, M., De Clercq, D., Breine, B. (2022). Foot strike determines the center of pressure behavior and affects impact severity in heel-toe running. Journal of Sports Sciences, 40(7), 808–820.
  42. Van der Meulen, L., Bonnaerens, S., Caekenberghe, I.V., Clercq, D., Segers, V., Fiers, P. (2024). Habitual running style matters: Duty factor, and not stride frequency, relates to loading magnitude. Journal of Human Kinetics, 94, 37–45.
  43. Van Hooren, B., Jukic, I., Cox, M., Frenken, K.G., Bautista, I., Moore, I.S. (2024). The relationship between running biomechanics and running economy: A systematic review and meta-analysis of observational studies. Sports Medicine, 54(5), 1269–1316.
  44. Van Middelaar, R.P., Reenalda, J. (2025). Estimating tibial bone load during running using only inertial measurement units on the tibias and sternum: A Centre of Pressure approach. Journal of Biomechanics, 189, 112839.
  45. Van Oeveren, B.T., de Ruiter, C.J., Beek, P.J., van Dieën, J.H. (2024). The biomechanics of running and running styles: A synthesis. Sports Biomechanics, 23(4), 516–554.
  46. Verheul, J., Hughes, O., Hitchens, L., Atherton, T., Sauter, T., Radwan, A., Pryse, E., Mitchell, A., Burton, S. (2026). Markerless motion capture for running: Validity and reliability of whole-body, joint, and muscle kinematics. Journal of Biomechanics, 195, 113133.
  47. Willwacher S, Kurz M, Robbin J, Thelen M, Hamill J, Kelly L, Mai P. (2022). Running-related biomechanical risk factors for overuse injuries in distance runners: A systematic review considering injury specificity and the potentials for future research. Sports Medicine, 52(8), 1863–1877.