kuleuvenneuromechanics / muscleredundancysolver Goto Github PK

INPUT

• OpenSim model
• List of inverse kinematics (.sto files)
• List of muscle fiber length data (.sto files)
• Optional: List of EMG (.sto files)

SETTINGS

• Parameters to estimate: only tendon stiffness -- only slack length & optimal fiber length -- all previous
• For which muscles to couple activations
• Type of coupling (waveform or maximal difference)
• Bounds on parameters (tendon stiffness and/or optimal fiber length)

CONSTRAINTS

• Track ID torques
• Bounds on parameters (provided)
• Coupling between muscles activations (waveform or maximal difference)
• Coupling between bounds optimal fiber length and bounds slack length (Antoine?)

COST FUNCTION

• Reserve actuators (normalized; /dof)
• EMG tracking ((act_exp - act_sim); /nMuscles_EMG)
• US tracking (normalise (lm_exp-lm_sim)/(lm_opt_generic); /nMuscles_tracked)
• Effort (activations; /nMuscles)
• Regularization

Not robust against estimating paramaters of muscles not encompassing the specified DOFs --> iM contains 0's

see: https://github.com/KULeuvenNeuromechanics/MuscleRedundancySolver/blob/master/solveMuscleRedundancy.m#L402

Now in the code we are using vMTtilde as vMtilde, and they are not equal if we consider the pennation angle.

Line 17 in HillModel_RigidTendon.m should be updated by:
vM = vMT.*cos_alpha; % here vT = 0
vMtilde = vM./lMopt;

Line 22 should be updated by:
[FMpas,FMactFL,FMactFV] = getForceLengthVelocityProperties(lMtilde,vMtilde,vMmax);

• Warning/error if not all provided names of dofs and muscles can be retrieved from the different motion files
• Time vectors of different motionfiles are not consistent

Hey, I have forked the repository so I can continue making additional changes without risking the main code.

So far, I have allowed pennation angle to be optimised as I think this may be the reason the fibre lengths are ending up approximiatly 6x normal. Currently carrying out testing.

In addition, I have attempted to solve the sqrt(-ve) problem by making lM.^2 - w.^2 a constraint that >0. Might not be the right solution but I was having issues before hand.

I have also had to increase lMtilde_max back to 1.8 as 1.5 was leading to a infeasable solution.

A few comments already:

• Bounds are not properly implemented for EMG [Solved]
• Change name of auxiliary variable (lM_projected?)
• we need some checks if the input data is compatible with the selected optimization variables. For example here: https://github.com/tomvanwouwe1992/MuscleTendonEstimator/blob/master/IO_Functions/useful_scripts/GetIndices_US.m#L14 [Solved]
• Clean up comments and delete unnecessary input arguments