GaitDynamics - Ground Reaction Force and Kinematics Prediction

GaitDynamics is a generative foundation model for rapid analysis of human walking and running biomechanics. This demo showcases its ability to estimate ground reaction forces from flexible combinations of kinematic data, representing Downstream Task 1 from the study: Publication | Github Repository

General instructions

This code is for ground reaction force and missing kinematics prediction using flexible combinations of OpenSim joint angles. The joint angles should meet the following criteria:

2. Resample to 100 Hz if you have a different sampling rate.

To use this code:

1. Upload OpenSim model (.osim) and kinematics (.mot) file. If kinematics are provided for all coordinates, the model will only predict ground reaction forces. If kinematics are missing, the model will also predict missing kinematics.
2. Enter the height and weight of the participant, and the treadmill speed (if applicable).
3. Click the "Submit " button.

Processing example data

1. Download the example files:
2. Upload data for either of the two example .mot files.
a. Predict ground reaction forces with a complete kinematics file by uploading the model file (.osim) and complete kinematics file (.mot).
b. Predict ground reaction forces with an incomplete kinematics file by uploading the model file (.osim) and the incomplete kinematics file (.mot).
3. Update the following input parameters.
a. Height = 1.83 m
b. Weight = 71.4 kg
c. Treadmill speed = 1.15 m/s
4. Click "Submit".
a. The example with complete kinematics should take a few seconds to complete.
b. The example with incomplete kinematics should take a few minutes to complete.
5. Plot the data, starting with GRF Results and choosing columns (e.g., force_r_vy).

Input Data and Parameters

Results

Visualization

Ground reaction force (GRF) results are available for all inputs. Missing kinematics results are only available if incomplete kinematics were used as inputs.
GRF column names follow these conventions:
• Starts with force_l or force_r: indicates the left or right foot, respectively
• Ends in _vx, _vy, _vz: indicates the magnitude value of the force in the x, y, and z directions, respectively.
• Ends in _px, _py, _pz: indicates the point location of the center of pressure in the x, y, and z directions, respectively.
• x is the anterior direction, y is the vertical direction, and z is the direction pointing to the right (medial or lateral direction depending on the leg).
Select output file
Column

Citation

Tan, T., Van Wouwe, T., Werling, K.F. et al. GaitDynamics: a generative foundation model for analyzing human walking and running. Nat. Biomed. Eng (2026).

https://doi.org/10.1038/s41551-025-01565-8