Muscle Recruitment

Hi,
AnyBody uses three different muscle models:
AnyMuscleModel
AnyMuscleModel3ELin
AnyMuscleModel3E
I think i have understood the difference between these three models and I’m using mainly the AnyMuscleModel and the AnyMuscleModel3E, as the results are mainly in a good range.
But I have a question regarding the optimization algorithm. Am I right with the assumption that I can choose for each of theses models one of the following algortihm in AnyBody?

  • Polynomial criteria
  • Soft saturation criteria
  • The min/max criterion
    As I know from the work of Forster, 2003, the results can be quite different between linear or a quadratic solver solution.
    Is my assumption right, that the min/max criterion is the simplex solver? If then, what is the addition for example MinMaxNRsimplex for?
    If I want to apply a quadratic algorithm I have to change the Penalty form LP to QP, but what influence has then the simplex solver?
    As I see mainly the MinMaxNRSimplex solver in the models of the repository, it is advisable to use a quadratic solver? Is it beacuse the simplex solver is more efficient and robust?
    I hope you can make this more clear to me.
    Best regards,
    Thomas

Hi Thomas,

Sorry for the late answer. First let me tell you that in the new version 4 of AnyBody the solvers are completely new and improved. They are much more robust with new algorithms and new possibilities.
There should be a webcast soon about it that we invite you to see.

So until version 3.1 the algorithm available are the Min/Max and quadratic criteria.
The solvers MinMaxSimplex and MinMaxNRSimplex are using the MinMax criteria. The difference is the robustness because they use a different library. They will always give the same result but where one fails to solve the problem the other may be able to solve it.
To use the quadratic criteria you can chage the penalty to QP but you also have to change the solver to MinMaxOOSolQP. The Qp penalty will have no effect on the simplex solver.
And the reason why we mostly use the MinMaxSimplex in the repository is both because it is more robust and because we believe it describe the real behaviour of the muscles better than the quadratic.

About the new solvers of version 4 there are 5 algorithms available now: MinMaxSrtict, MinMaxAux, Linear, Quadratic and Polinomial. So there are now more possibilities and the rubustness has been improved also. Keep an eye on the coming webcasts for more details on the new solvers.

I hope it makes it more clear, please writte again if you need more info.

Best regards, Sylvain.