Inverse Dynamics Problem in gait analysis


It’s a while that I changed the default “GaitLowerExtremity” demo file for AnyBody version 5. 0. 0. 2014 according to our lab settings and needs (attached zip file). Also, my concerns and applied changes to the original model are described in more detail in “_Questions.docx” inside this zip file. In brief, big joint one directional joint moments are introduced for SacrumPelvis and spinal processes. I would appreciate any comments/suggestions in this regard that could help me solving this issue.


  1. Inverse dynamics results are checked. For example: “Main.Studies.InverseDynamicStudy.Output.HumanModel.SelectedOutput.Right.Leg.JointMomentMeasure.AnklePlantarFlexion” shows reasonable values.
    However, the outcomes for joint moments in spinal processes are vague. In particular it is expected to observe moments that are roughly oscillating around zero for all directions, i.e., Axial, Lateral, and Extension moments. Actually, the same concerns exist for SacrumPelvis moments, therefore transferring to spinal processes. These joint moment values are reasonable for the original Demo file (i.e., GaitLowerExtremity), however, for our model, axial and lateral moments are showing big one directional moment! There are some periodic changes, however, as a result of a big shift in values, there are some serious issues. I could not find the reason and appreciate any help in this regard. The same concern occurred for other participants.
  2. I did not change any other coordinate systems, other than the ones described below. Should I change in other code parts as well? More generally, what other concerns/changes are required/suggested.
  3. Could the above discrepancies occur as a result of inaccurate anthropometric data? In detail procedure for anthropometric measurements used in “TrialSpecificData.any” is not provided.

Model Descriptions:

  1. This model is based on the default “GaitLowerExtremity” demo file for AnyBody version 5. 0. 0. 2014.
  2. Model folder is placed on ‘Example’ folder containing all initial demo files:
    …\AnyBody Technology\AnyBody.5.x.x\AMMRV1.3\Application\Examples
  3. I did not incorporate all data files in an integrated C3D file. Therefore, the corresponding force plate data is used in TSV format, named “FP.TSV”. Therefore, the force plates are defined in customized “ForcePlateType4.any”, used along with other files in attached folder.
  4. The force plate definition and geometry are defined in “InfoFolder.any”, including CAL_MATRIX, reading the FP.TSV file, filtering the force data, and assigning data from “FP.TSV” to relevant force and moments for force platforms.
  5. “Environment.any” file also changed slightly to introduce the relevant force columns to variables used in AnyBody calculations.
  6. Main difference is the coordinate systems, where global coordinate in default demo file is changed as follows: +X-X, +Y +Z, +Z+Y
    Therefore, in “TrialSpecificData.any”, PlevisRot variables are changed accordingly.
  7. It is also notable that the local coordinate system for force platforms used in our model has been changed from local coordinate system used in original demo file, i.e.,
    +y-x, +x+y, +z+z

Dear Ehsan,

One way of checking such a model would be to run it and look at the residual forces present between human and world. In the gait models this force is found in the pelvis driver, if the forces there are very big it is a sign that the external forces are not applied correctly.


You have high loads in Y and Z direction, so there seems something wrong. Let me know if you found out more.