simtk.org

Hello,

I've been working on improving the results of my static optimization (run through the MATLAB API).

I'm able to get a stable result when I feed the controls back into forward dynamics, but there are still some errors when compared with the original inverse kinematics.

I've added residual/reserve actuators which has helped, but are there any other techniques to improve the accuracy? Is it possible to weight kinematic accuracy more than minimizing muscle activations or is it likely to be issues with the model? If I've already added residual actuators, are there other ways I can improve the model?

Thank you!

Static optimization does not ensure continuity between time points analyzed and uses a rigid tendon assumption, so the results from SO will not be able to drive a forward simulation stably without a controller.

If you need to be able to do this, consider some other methods: https://simtk-confluence.stanford.edu:8 ... +Workflows

In particular, CMC, direct collocation, or shooting methods may work for you, depending on your needs.