Simulation of robot with rubber chain drive

[Gickl]

Hi,

i have to virtualize the Jaguar V4 robot platform (see picture) to simulate its behaviour.
but i am not sure what the best way is to simulate the rubber chains. it might be possible to build the chains of many small parts so that they are flexible. or maybe it is better to use softbodies, i don't know...

so if someone has an idea please tell me! it should be as easy as possible but also as realistic as possible!

Thank you in advance
Gickl

[Basroil]

Well, with that type of robot you have a set of issues from continuous collisions to concave deflections in the section between the wheels. The main body is just a box, so that's easy enough, the problem is going to be the tracks.

If you ignore those and don't mind making custom collision functions for the tracks, you could use a convex shape for each part and do track rotation characteristics in the custom collision (basically anisotropic friction where the friction direction parallel with the track rotation is done manually).

If you're less adventurous and more into joint chains ( and don't mind the headaches that will come from that), you could also model the tracks as a series of wheels attached by springs that rotate in unison. You'll probably run into some odd behavior though, chains don't tend to work exceptionally well if they are too long.

Then you also have the non-bullet (and probably most recommended one) where you use your own physics calculations and just use the collision system of bullet to help refine your physics.

I've tried a few different non-wheeled robots in bullet, and I can tell you physical accuracy is one hell of a task.

[kalesony]

I also don't think you can expect much "physical" accuracy, especially when considering friction. "Gaming" physics engines are aimed at visual plausibility and "interactive" frame rates, so you will probably get it to work for visualization purposes, but precise simulation with results comparative to a real-world experiment will be very hard to achieve (in my opinion).

Anyway, if it is mainly a visualization and you plan on doing this using bullet - let me know: maybe we could cooperate since a similar task is hanging over my head (but chances are slim since I am not sure if I have time to do that in the near future).

Cheers,
kalesony

[Gickl]

i just got the robot to work alright. it is not perfect but resonable. i used 4 btCompoundShapes consisting of a btCylinderShape and btBoxShapes as lamellae. to get a flat underside i use a btBoxShape and a CustomMaterialCombinerCallback which applies custom friction (see http://www.bulletphysics.org/Bullet/php ... f=9&t=6076). So the main tracks look like in the picture below. i use the same concept for the flippers and the whole robot works almost like in reality.

[Basroil]

i just got the robot to work alright. it is not perfect but resonable. i used 4 btCompoundShapes consisting of a btCylinderShape and btBoxShapes as lamellae. to get a flat underside i use a btBoxShape and a CustomMaterialCombinerCallback which applies custom friction (see http://www.bulletphysics.org/Bullet/php ... f=9&t=6076). So the main tracks look like in the picture below. i use the same concept for the flippers and the whole robot works almost like in reality.

Sounds like you managed to find your way around that bot!

Are we going to be able to see some cool videos soon?

[Gickl]

i'm not sure already if i am allowed to show videos of the robot and what my professor at the university thinks about that, respectively

but i will see what i can do