simtk.org

I am testing the version 1.8.1 with the gait2354 model and the example files. I was wandering if it is possible, to do the cmc computation and the forward dynamics for a whole gait cycle. I have done the CMC for the first 5 seconds. The excitations look fine and reasonable, but when I now start a forward dynamics simulation, the legs start "turning and twisting" at different specific times, e.g. I can simulate from 0.8 to 1.2 (like in the original example) but if it comes to 1.5 the simulation become unreasonable and finally breaks down.

Is anyone here who encountered similar problems or has any ideas for me?
\Nicolas

Hi Nicolas,

The issue is primarily numerical. When running forward with CMC result, the simulation still diverges from the actual CMC run and overtime this causes the simulation to blow up since forward is run in Open-Loop (while CMC is run as a controller with feedback).

This depends however on the dataset, filtering, and a bunch of other factors and there have been simulations of more than one gait cycle using CMC results.

The next release should allow users to do this more robustly.

-Ayman

Dear Ayman,

thanks for the response. What you have said may solve my problems, if I do long forward dynamics simulations in time.

But besides that I have a different problem: my simulations (forward dynamics) become unreasonable and break down at specific parts of the motion (in my test cases always between the stance and the swing phase). What I mean is: even if I limit my simulations to, let's say 0.5 seconds, I can not get reasonable results, if in the motion such a change (stance->swing) phase occurs.

Do you have access to one of those simulations which have simulated more than one gait cycle using the cmc results and could you send them to me?

Thanks,
\Nicolas

*update:

sorry, but during my latest test I have found out, that the unreasonable results not only occur between the stance and the swing phase.

Furthermore I found that it is not possible to start the simulation later than 1.7sec. Could this point to a problem in the CMC results? I have looked on several muscle activities and to me, most of them seem reasonable and periodical, even though there are some unreasonable peaks and some noise in some points of time.