MinMaxNRSimplex vs. MinMaxOOQpSolver

Thank you for helping me with my muscle wrapping problem - the model
is now up and running. However I have run into another problem. I
realized that the forces that I was applying to each hand were
actually the forces that were applied to the wheels (so to fix that I
essentially negated the forces applied to the wheels as the forces in
the hands are equal and opposite). When I did this to one side to
test it, the model would no longer run because the muscle recruitment
analysis failed. I then played around with my Study section and
changed the MinMaxNRSimplex to MinMaxOOQpSolver and the model was able
to run, however the resulting joint forces were unexpected (and I
think, incorrect). This was the only thing I changed and I was
wondering what the differences in the solvers are - especially since
my model will run with one solver, but not the other.

Also, when I put my forces into the model, the are nice and smooth,
however when I run the analysis, joint forces appear jagged. Is there
any way to smooth this out? Add more nSteps?

Thanks for all your help,
Sarah

Hi Sarah

The solution which will come out of the two solvers will be the
same, if you use the default settings for them, ie. no penalties.
Since the NRSimplex slover will not recruit your model it seems that
the model is difficult to solve. The OOQpSolver solver which solves
the recruitment uses a different algorithm which essentially will
give the same results as the NRSimplex (with default settings) but
since the implementation is different you may see cases which can be
solved by one of the solvers but not the other one. This is due to
numeric round off errors.

It is difficult to tell what could be the reason for the solving
problems. In the new repository 6 we have made a small change in the
glove segment we use to get good boundary conditions for the hands.
This change will in some cases solve recruitment problems.

http://www.anybody.aau.dk/repository/Repository6/index.htm

Please find the file at BRep/Aalborg/Arm3d/Glove.any and rename your
current glove file and replace it with this one, it migth help solve
the problem.

Finally problems may also arise from the the acceleration forces; i
recall that the model is driven by marker data, so please check for
example the rrdot for the hands to see if the acceleration of the
hands looks ok. It is a typical problem that the acceleration forces
for marker driven model can become too large.

Please write again if you have further questions

Best regards
Søren, AnyBody Support

— In anyscript@yahoogroups.com, “Sarah R. Sullivan” <sarsulli@…>
wrote:
>
> Thank you for helping me with my muscle wrapping problem - the
model
> is now up and running. However I have run into another problem. I
> realized that the forces that I was applying to each hand were
> actually the forces that were applied to the wheels (so to fix
that I
> essentially negated the forces applied to the wheels as the forces
in
> the hands are equal and opposite). When I did this to one side to
> test it, the model would no longer run because the muscle
recruitment
> analysis failed. I then played around with my Study section and
> changed the MinMaxNRSimplex to MinMaxOOQpSolver and the model was
able
> to run, however the resulting joint forces were unexpected (and I
> think, incorrect). This was the only thing I changed and I was
> wondering what the differences in the solvers are - especially
since
> my model will run with one solver, but not the other.
>
> Also, when I put my forces into the model, the are nice and smooth,
> however when I run the analysis, joint forces appear jagged. Is
there
> any way to smooth this out? Add more nSteps?
>
> Thanks for all your help,
> Sarah
>