# Optimization algorithm

Hi,

What kind of optimization approach do you use in the software (forward integration, static optimization, direct collocation, …) ?

I cannot find the answer in the threads, in the software documentation or in the published papers. In the paper of Rasmussen (2001) Muscle recruitment by the min/max criterion … , it is mentionned: “the feasible direction method (Vanderplaats, 1984) was used to solve the non-linear optimization problem”. Was it changed since this paper was published? Is it still the case with v. 5.3 of the software ?

Thank you.

Pierre

Dear Pierre,

This is not my strong side and vacation time has reached us here, so it will take a while (three weeks) before the right people is back.

There are many papers dealing with this topic did you see this paper/ abstract:

Rasmussen, J., Tørholm Christensen, S., Damsgaard, M. & de Zee, M. (2004), “The role of mechanics and optimization in ergonomics”, 5th ASMO-UK / ISSMO conference on Engineering Design Optimization, Stratford upon Avon, July 12-13, 2004, pp. 10 pp… [PDF]

In AMS 4.0 there was a major update to the recruitment solvers please have a look here.
http://www.anybodytech.com/index.php?id=116&tx_ttnews[tt_news]=53&cHash=e52c5ae828e34381e92de12c4b3980ff

I am sorry but this is what I can do at this point.

Best regards
Søren

Hi Soren,

Thank you for your reply. From what I understand, the min/max problem is converted to a linear problem with the bound formulation. Do you know which LP algorithm is then used to solve the LP problem (simplex, interior-point, etc.) ? In Rasmussen 2004, there is a paragraph that says:

The min/max criterion only concerns the muscles that are maximally activated or which help support muscles that are maximally activated. The system may contain muscles that have no influence on the maximum muscle activity in the system, and they are left undetermined by the problem formulation above. This means that the problem must be solved iteratively, where each iteration eliminates muscles that are uniquely determined and removes their contribution to the support of the external load from the right hand side. The next iteration then determines the submaximal muscles. This process continues until there are no muscles left in the system.

Is this the general approach used in AnyBody ?

Thank you.

Pierre

Hi Pierre,

Please also have a look in these papers:
http://www.sciencedirect.com/science/article/pii/S1569190X06000554

Rasmussen, J., Damsgaard, M., Tørholm Christensen, S. & de Zee, M. (2005), “AnyBody - decoding the human musculoskeletal system by computational mechanics”, Skallerud, B. & Andersson, H.I. (Ed): Konferanse i beregningsorientert mekanikk (computational mechanics). Trondheim, Norway, pp. 43-59. [PDF]

I think it contains some of the info you are requesting.

Concerning the iterative approach this applies to the MinMaxStrict criteria this is what is used in the papers above, please also see

Best regards
Søren

Hi Soren,

Thank you for the papers. I used the MinMaxStrict in my study.

Pierre

Hi Soren,

What about the kinematic solver ? For example, if I use soft constraints in my model, how these soft constraints are solved, with respect to hard constraints ? Which algorithmic approach is used ?

Thank you.

Pierre

Dear Pierre,

Andersen, Michael Skipper, Michael Damsgaard, and John Rasmussen. “Kinematic analysis of over-determinate biomechanical systems.” Computer Methods in Biomechanics and Biomedical Engineering 12.4 (2009): 371-384.
( http://www.tandfonline.com/doi/abs/10.1080/10255840802459412 )

Best regards,
Moonki