GaitLowerExtremity & FDK

I did my research with the GaitLowerExtremity model incorporated with FDK, but a problem occurred …
1.0) Inverse dynamic analysis…
Failed to resolve force-dependent kinematic constraints. Newton relaxation too small. (final force error = 2.995027E+001)
Constraint no. 0 above error tolerance 20.000000, error = 29.950273.
Constraint no. 1 above error tolerance 20.000000, error = 27.179228.
1.10) …Inverse dynamic analysis terminated
ERROR(OBJ.MCH.KIN3) : D:/a…s/AMMR/A…n/E…s/G…y/GaitLowerExtremity.main.any : InverseDynamicStudy.InverseDynamics : Kinematic analysis failed in time step 10
And my settings related with the tolerance are :
InverseDynamics.ForceDepKinOnOff=On; //Values gained by experience
InverseDynamics.ForceDepKin.ForceTol=20; //Values gained by experience
InverseDynamics.ForceDepKin.Perturbation=5e-5; //Values gained by experience
InverseDynamics.ForceDepKin.LocalSearchOnOff = Off; //Values gained by experience
InverseDynamics.ForceDepKin.MaxNewtonStep = 10; //Values gained by experience
can you help explain this error?
Thanks so much.

Hello Thomas,

the error you posted shows up when the because the kinematics could not be solved while the FDK solver tried to solve the force equilibrium for the FDK degrees of freedom. This could happen when the system has to solve a bad position during the solution of the FDK system for finding a force equilibrium.

In general, it is difficult to tell what could be the best way to solve this problem since there could be different strategies. One way could be to add artificial springs to the system to kind of help the FDK solver.

Which DOF did you use for your FDK model?

Best regards

Dear Daniel,
Thanks so much for your reply!
I used three dofs of the hip joint and the springs had been already defined in the library model .
Maybe the C3D data is not good, so the motion can’t be resolved properly.
And I have another question, does the initial position influence the procedure of solving the problem?
Best regards

Hi Thomas,

I think it could be a bit tricky to use a model based on C3D data with FDK since the motion is optimized to fit the marker positions and due to an FDK joint, positions of the segments might change a bit which could be a problem for the kinematics. But it is definitely worth to try.

In general the initial conditions can make a difference. A good initial guess can change a lot in terms of speed and quality of the results.

Best regards

1 Like