btMultiBody simulation instability / explosion of energy

[erwincoumans]

In some cases the new btMultiBody simulation is unstable, and gains a lot of energy. This needs to be tracked down and fixed. See also http://bulletphysics.org/Bullet/phpBB3/viewtopic.php?f=9&t=10237

[donnyward]

I've modified the MultiBodyVehicle demo to use my btMultiBody construction code. At the top of MultiBodyVehicle.cpp, you can modify two constants (FIXED_STEP, NUM_MULTIBODIES) to change the time step used, or change the number of btMultiBody in the simulation.

https://gist.github.com/donnyward/f2130d62915bd8b896cf

[erwincoumans]

That's great, thanks for the reproduction case. I am tracking down the issue now.

[erwincoumans]

I tracked down some causes of instability. One of them is self-collision can cause near-singular configurations where resolving the collision adds a lot of energy. We can add a fix/hack for this, but I need to figure out if that is ok to do. Another potential instability is rotational gain, probably due to coriolis forces, but it could be a bug too. Enabling some damping reduces/hides this issue. I'll post again when I find out more.

[erwincoumans]

I fixed some issue that was leading to the self-collision 'explosion', see #295
There is still an issue of energy gain, likely due to explicit integration of coriolis forces, so damping is still needed for each multibody add:
multiBody->setLinearDamping(0.1f);
multiBody->setAngularDamping(0.9f);
Please open a new issue if there are still energy gain/explosions with btMultibody when using damping.