The collision detection for the cylinder geometry fails on the ground plane and the simulation overcompensates, causing this twitching behavior. I believe that this will occur frequently when cylinder geometries are vertical on the ground plane, but I haven't setup a basic test case outside my own testbed. I also think this is just inherent to the Bullet physics engine.
I believe that this can be fixed with putting another geometry (such as a sphere) on the ends of the rods, and I will update this issue if that is the case. We would like to have a way to address this problem long term however.
EDIT: Performing a test on a tgBoxGround instead of tgPlaneGround seems to fix this issue in the short term. I no longer see the same compensating jumps on the cylinders for similar movements. The same issue still occurs for tgPlaneGround however.
I have recently rediscovered a potential simulation issue with cylinders and the ground plane.
In the following video (within the first 15 seconds), you can see the issue:
https://drive.google.com/open?id=0B-zbEExDXevaWWwzdWhzejlPTkE
The collision detection for the cylinder geometry fails on the ground plane and the simulation overcompensates, causing this twitching behavior. I believe that this will occur frequently when cylinder geometries are vertical on the ground plane, but I haven't setup a basic test case outside my own testbed. I also think this is just inherent to the Bullet physics engine.
I believe that this can be fixed with putting another geometry (such as a sphere) on the ends of the rods, and I will update this issue if that is the case. We would like to have a way to address this problem long term however.
EDIT: Performing a test on a tgBoxGround instead of tgPlaneGround seems to fix this issue in the short term. I no longer see the same compensating jumps on the cylinders for similar movements. The same issue still occurs for tgPlaneGround however.