Making a rope / chain

[darwinian]

I'm attempting to create a rope/chain object, but it is not behaving as I expected. I have a series of small bodies linked together by joints. Now if I hold one end of the rope fast, and apply a force to the opposite end to straighten it out, the bodies move apart and the joints appear to stretch between the them, rather than stay rigid as I would have expected. I have tried using pin joints and slide joints, and both suffer the same problem.

Any ideas what I could be doing wrong? Cheers.

[X-0ut]

The documentation says slide joints are good for modelling a chain, but I got it working fine using pivots.
Basically I just had circle shapes for links with a body for each shape, and a pivot between each.
The pivot point goes on one of the links.

[slembcke]

You're running up against the limitations of the constraint solver. The more of the following you do, the better:

• Use more solver iterations.
• Use a smaller timestep.
• Put less force on the chain.
• Put fewer bodies in the chain.
• Use a constant timestep.

The first two items increase the CPU usage, the last two decrease it. You'll have to find a balance. If you aren't using a constant sized timestep, you really should be.

[citrys]

What's the practical difference between increasing the number of iterations for the solver (in Ruby, space.iterations) vs. stepping the space more frequently with a smaller timestep?

[slembcke]

What's the practical difference between increasing the number of iterations for the solver (in Ruby, space.iterations) vs. stepping the space more frequently with a smaller timestep?

That's a good question without a really good answer I think. Doubling the iterations or halving the timestep will not double the quality of the solution. Also, using a coarse timestep with a ton of iterations (or vice versa) doesn't really help either. In my experience, you have to find a good balance by experimenting.

[darwinian]

Thanks for the responses.

After some more experimentation, I'm still not really able find an adequate solution. I am using a constant timestep of 180 steps per second, with ten solver iterations.

With only one joint, ie a rigid pole, the joint behaves fine, with a body of any mass swinging around on the end of it. As soon as an intermediate body is added (so that there are now two joints between the anchor point and the suspended mass), the joints will stretch. A solution to this is to reduce the mass of the suspended body, however it seems that the mass has to be reduced significantly (to roughly < 0.1) in order for the joint not to stretch.

Does this seem reasonable? I would have expected larger masses to still behave correctly.

Thanks again.

[slembcke]

Ahh. So you are making a chain of really light objects and putting a heavy object on the end? That is a recipe for disaster in most physics libraries. Same problem with letting a really heavy object land on a light one. Because the light object has so little effect on the heavy one, it takes a lot of solver iterations to get a decent solution.

What I would recommend in this case is to even out the masses, or to use a single slide joint from the object to the anchor point and simulate the rope using vertlet physics. In the latter solution, the chain wouldn't be interactive, but it would look pretty at least.

[dieterweb]

Hi,

Has anybody implemented the approach with vertlet physics for drawing a rope? It would be great if the one would share some code for this.

Thomas

[slembcke]

The code for simulating the physics (http://www.teknikus.dk/tj/gdc2001.htm) or code for drawing the rope?

[dieterweb]

I already implemented it by myself.
But thanks for the link. Very useful site.


Thomas