Joint coming apart

[nikolaus]

I have a robot with a 3-dof hip joint, using three revolute joints.
When I start the simulation, the robot's body pitches forward, not because the pitch joint changes position but because the hip roll joint comes apart. The blue joint core separates from the orange outside cylinder. Am I doing something horribly wrong? This happens in my ankle, as well. The only common thing I can see is that in both cases there are joint with intersecting axes of rotation.
Thanks!

[coppelia]

Hello Nikolaus,

I see two potential causes for that to happen:

1. The links are overlapping and colliding. Make sure they do not overlap or that they have their respondable masks set correctly.

2. The links have too small masses, or a too small moment of inertia. Or the proportion between masses involved is greater than 10. Refer also to the design considerations here.

Cheers

[nikolaus]

I've been thinking about how to redesign our robot model (it looks like our problem is the link-mass proportion), and I've got a follow-up question: if you were trying to make a universal joint between two large masses, how would you do it? In real life, this uses a small-mass link between two large masses, and so that's what I modelled in V-REP, but clearly it's not working. :)

This page shows three revolute joints directly in series -- is this actually possible to simulate? I've tried putting joints in series like this and they get exclamation marks next to them during simulation and fall apart. If we could do this, though, we could just get rid of our small mass entirely.

Thanks!

[coppelia]

Hello Nikolaus,

You are using dynamics (i.e. the physics engines), so this puts some restrictions on the way how dynamic elements have to be build in the scene hierarchy. This includes that two joints always need a link in-between (i.e. a shape that is non-static).

Well, if your spherical joint construction is not actuated (i.e. passive), then you could directly use a spherical joint type.

But if it is actuated, then I would simply add spherical shapes between the revolute joints. The total mass of the construction should be what the real robot has, but you can evenly balance masses (so that each element of the spherical joint construction weighs the same (and with a similar inertia)). Refer to the model "Asti", that has also an actuated spherical joint construction for the hip.

Cheers

[nikolaus]

The joints are actuated, so a spherical joint is unfortunately out of the question.
Thanks for the tip about Asti—I'll try to replicate that setup. Just gotta do some math to balance out the masses and rotational inertias.