CMC Time Steps extremely small

[Kevin Hao]

Recently, the timesteps during CMC have been extremely small. The time steps in my Inverse Kinematics file are at 0.01s intervals and previously, CMC always used those same steps, but now, it will step by 0.00001 or a something similarly small. Adjusting the integrator settings for the minimum time step does make it larger, but the algorithm is still defaulting to the minimum value. Any ideas why and what I can do to fix it? Thanks!

[Thomas Uchida]

it will step by 0.00001 or a something similarly small

Relatively small integration time steps can occur if the system contains relatively high-frequency dynamics. Possible culprits include short tendons, spiky input data, bad simulation settings, bad contact model parameters, and large forces being applied to bodies with small masses. Also see the "Best Practices and Troubleshooting Tips" section on the "Getting Started with Forward Dynamics" page in the Confluence documentation (https://simtk-confluence.stanford.edu/d ... ootingTips).

Adjusting the integrator settings for the minimum time step does make it larger, but the algorithm is still defaulting to the minimum value.

Forcing the time steps to be larger will increase the integration tolerance by an unpredictable amount; instead, you could set the integration tolerance to a higher value. However, it might be best to track down the culprit (e.g., by plotting muscle states, kinematics, GRFs, etc.).