Plane perturbations

[kate]

Hi,

I've just updated to the latest version of Bullet, and found that the recent change to the btConvexPlaneCollisionAlgorithm (made in r1650, "disable perturbation for btConvexPlaneCollisionAlgorithm, it introduces artifacts (additional points cause spheres/cylinders and other curved objects to move)") caused box shapes to jitter about on a static plane.

Setting the values of m_numPerturbationIterations and m_minimumPointsPerturbationThreshold back to 3 fixes this for me; I was just wondering if anyone else found the same thing happening and/or if there's a better way to get round it?

Regards,

Kate

[Erwin Coumans]

caused box shapes to jitter about on a static plane.

Is this jitter while the box is at rest on a btStaticPlaneShape?

Can you reproduce it in a Bullet demo, with the same collision shapes, sizes, timestep, collision margin etc. that you use?
Thanks a lot,
Erwin

[kate]

Hi Erwin,

Is this jitter while the box is at rest on a btStaticPlaneShape?

It's when the box falls onto the plane (objects that start off resting on the plane seem OK, perhaps due to being decativated). I've made a video (attached); the box starts at a small height above the plane and drops.

Can you reproduce it in a Bullet demo, with the same collision shapes, sizes, timestep, collision margin etc. that you use?

OK... I need to fix another issue first, but once that's done I'll look at making a demo to reproduce this. It seems to be only box shapes that are affected (cuboid convex hulls seem OK); the boxes penetrate the plane a little, which is presumably what causes the jitter.

I think it might be due to the small size of the objects - the cube in the video has edges 10cm long and a collision margin of 0.004; if I scale both values up by 5 or 10 they don't jitter. Identical small cubes were fine in previous versions of Bullet, though... I'm happy to just set the perturbation values back to 3 to keep backwards compatibilty, if it's not something that's affecting other people (it could just be my set-up, I guess...).

Thanks,

Kate