Simscape: The derivative of position in a direction not equal to the velocity

I am using simmechanics 2g to simulate a robot. On one of the points on the robot, the velocity (plotted using scope) in the horizontal-x direction does not match the time-derivative of the position. This seems to defy my understanding of Newtonian Mechanics. Has somebody faced similar situations or know why this problem arises?

Note: There are no external forces provided in the horizontal direction although there is one provided in the vertical direction.

[Rebecca Krosnick]

What equations and blocks are you using to generate the time-derivative of the position and the velocity? Can you also provide screenshots of the scope plots?

[Sarvesh Srinivasan]

Here's how the model looks (the three-port subsystems have been set to zero, thus having no external force in the x-direction). The gravity's in the y direction. The scope on the left (scope3) calculates the difference between v_x and the time derivative of x.

The scope graphs this out. Despite it being less 1, it's still a significant difference.

I would be thankful for any explanations given for this observation.

Answer 1

[Sebastian Castro]

In general, I wouldn't trust the Derivative block's results very much. This block only does an approximate derivative of a signal, which is highly dependent on the simulation step-size taken, and has no continuous states or error control.

Some useful links:

• Derivative block documentation
• This MATLAB Answer

Try doing it the opposite way: Use an Integrator block on the velocity signal and compare it to the position. This will likely look a lot better.

- Sebastian

[Sarvesh Srinivasan]

Changed the model as you suggested and tried it as follows

The scope gives this result. The output looks less erratic but there is still a significant difference varying with time. Could you explain this?

[Sebastian Castro]

Could be the solver -- I would guess this especially since your results are "jagged". The default "ode45" solver isn't always ideal for SimMechanics models.

If you go to your model's configuration parameters and switch the solver to ode23t, how do things look?

- Sebastian

[Sarvesh Srinivasan]

Thanks for the suggestion, but the results are the same regardless of the ode solver I use, and this includes even fixed-step solvers. Any other suggestions to proceed?

Also, thanks for helping through this even in the holiday season. Really appreciate it :)

[Sebastian Castro]

Hmm... It could also be the Max step size parameter in the same solver settings. You could try reducing it and seeing what happens.

- Sebastian