OpenGRF: Predicting Ground Reaction Forces and Moments During Daily Living Activities in OpenSim

Abstract

Background and ObjectivesGround reaction forces and moments (GRF&Ms), typically measured using force plates, are key inputs for musculoskeletal simulations. OpenSim currently lacks a tool to predict GRF&Ms directly from kinematics. This study presents OpenGRF, an OpenSim-based tool designed to estimate GRF&Ms and the centre of pressure (CoP) from joint kinematic data and validates its performance against force plate recordings. MethodsThe proposed methodology integrates calibrated foot-ground contact probes with an optimization framework based on computed muscle control, while CoP is computed accounting for both kinematic and dynamic contributions. For validation, a scaled FullBodyModel (37-DoF without muscles) was created for seven healthy adults performing six trials each of level walking, stair ascent, and stair descent, for a total of 126 marker-based trials. GRF&Ms predictions were compared to reference force plate data using normalized RMSE (nRMSE), Pearson correlation coefficients ({rho}), CoP error, and Statistical Parametric Mapping (SPM) ResultsResults showed high accuracy for vertical GRF (nRMSE [≤]1.5%, {rho} [≥]0.94), particularly during level walking, and good accuracy for anterior-posterior GRF (nRMSE 4.4-6.1%, {rho} = 0.81-0.91). Medio-lateral GRF was less reliable, especially in stair tasks (nRMSE up to 11.2%, {rho} down to 0.48). Free moments were the most challenging quantity to predict across all tasks (nRMSE up to 28%). In contrast, ankle moments were predicted with high fidelity (nRMSE {approx}1.7%, {rho} {approx}0.98). Median CoP errors were 21-23 mm, with largest discrepancies during double support. ConclusionsOpenGRF enables physics-consistent estimation of GRF&Ms and CoP directly from kinematics, achieving the highest accuracy for vertical GRFs and predicting ankle moments that closely match those obtained from force plate measurements.