Forces not working

[tsreyas]

When I create a model and simulate it in SOFA, for some models none of the forces applied activate at all. It might be due to overload, but time passes as usual and the framerate is at a value that should still allow me to see motion. What could be the possible causes for this?
I have attached a screenshot of one of the models during simulation, and the corresponding code.

code.txt

The blue cuboid is supposed to fall onto the legs, but it doesn't move at all.

[alxbilger]

What forces are you talking about?
If you talk about ConstantForceField, its total force is set to 0, so no force applied
If you talk about the constraints, I think you need to add a ConstraintCorrection component to each MechanicalObject. See https://www.sofa-framework.org/community/doc/simulation-principles/constraint/lagrange-constraint/#constraintcorrection.

[tsreyas]

Currently ConstantForceField is set to 0, but there is a gravitational force of -9.81 in the Z direction defined at the beginning, which refuses to work as well.
I'll check if adding a ConstraintCorrection to the model fixes it.

[hugtalbot]

Welcome @Hellscythe789

@alxbilger is definitely right, a ConstraintCorrection is missing, the list of possible ConstraintCorrections is available online.

As a reminder, when choosing a contact response "FrictionContactConstraint", you will generate constraints for each contact points (output of the detection). This requires therefore to be able to solve such constraints: this can be done using:

• a FreeMotionAnimationLoop
• a ConstraintSolver (the generic works for most cases)
• a ConstraintCorrection for each simulated object

The FreeMotionAnimationLoop computes four steps:

• free motion : objects can move freely, even going through another
• collision detection
• constraint resolution (constraint solver)
• correction using the contact forces: this last part cannot be done if the ConstraintCorrection is not present in the node

I hope this clarifies things.
Best,

Hugo

[tsreyas]

Sorry for the late update, but I've tried adding constraints and had the following results:

UncoupledConstraintCorrection:

LinearSolverConstraintCorrection and GenericConstraintCorrection:

I used the SparseLDLSolver for each of these.
After showing these, the program stops responding.
There wasn't any debug info for either case in the terminal either.
Is there something else going on?

PS: I'm still new to this, so please forgive me if my questions seem too elementary

[hugtalbot]

Hard to guide you more @Hellscythe789 without seeing and testing the simulation. Is there a repo where we can access it (with associated meshes) ?

Else, I would say maybe a problem of convergence: you could separately try increasing the number of steps in the ConstraintSolver, reducing the time step. Is the simulation returning any error in the console?

[tsreyas]

@hugtalbot I've tried increasing the number of steps in the constraint solver and decreasing the time step as well, but no go. The console doesn't show any errors either. I've created a private repo with the relevant mesh files and code, so you could get the files from there.

[hugtalbot]

I will give it a try @Hellscythe789

[hugtalbot]

From what I see, the problem of your simulation might be related to:

• mainly the fact that objects are ALREADY AT FIRST STEP in collision (e.g. legs and pads) which is not solvable you have to make sure objects are not colliding (regarding your collision settings like contactDistance="0.25") at start
• the absence of boundary condition like FixedConstraint or PartialFixedConstraint

Does this make sense @Hellscythe789 ?

[hugtalbot]

Removing the pads and adding <FixedConstraint indices="17 50 25 13" /> for each Ext ➡️ some kind of work (apart from the physical properties)

[hugtalbot]

any update @Hellscythe789 ?

[tsreyas]

Oh I'm so sorry, I didn't get a notif for these until now. I'll test these suggestions out, thank you so much, Hugo!

[tsreyas]

@hugtalbot you were right, the objects were already in the first collision step, so it was behaving oddly. I shifted the Ext's a bit upward and everything started working normally.
Although I wonder why it didn't show me an error message...
Anyway, thank you so much!

[hugtalbot]

Great to hear @Hellscythe789. Having non-intersecting objects is a condition in all physics engine, but I agree: although such circumstances might be not straightforward to detect, it would be nice to have a message warning the user! Would you mind create an issue about this?