Force vs. motion actuation in Simscape Mulitbody

Hi everybody,

I have some troubles with the Simscape Multibody Toolbox and hope that someone can help me.

For demonstration I built a small script and two Simulink-Models which are attached. Basically the script does the following:

• Calculate a rest-to-rest-motion for a crankshaft from 0 to pi/2
• Use this calculated motion in the simulink-model for simulating the necessary torque
• Use this torque to make the same simulation, but now actuate the model with the torque and not with the given motion
• Finally compare the originally calculated motion with the now simulated motion which you get from the torque actuated model.

The results are plotted here:

As you can see especially the simulated velocity differs from the originally calculated motion. I worked with a fixed-step-solver and a time discretization of dt = 1e-3 s, but the change of discretization has no effect to the described phenomena.

For a better replicabilty a attach the necessary files and models and hope that someone has an explanation for this.

Thank you very much!

[ENRICO]

I'm experiencing a similar issue with a simple mechanical system of some degree of freedom. I guess there's no significative chance to get same response using actuation input in place of motion imposed. Indeed inverse dynamics problem, while imposing whole kinmetics (speed and position and accelerations), or even computing them numerically from position, is a process by far more straight ans simple than solving model dynamics in a direc dynamic problem, requiring numerical integration of motion equations and mass matrix evaluation. Numerical errors cumulates quickly and there's little chance of not seeing them diverge quickly, even taking small integration steps down to 1e-6, making anyway simulation time impracticable, in absence of a control loop for keeping error reasonablly limited