Bouncing ball

[PhilB]

Hi experts,
I am using JBullet Release 20080917.
I've created a simple scene with a bouncing sphere. I've set the restitution coefficient to 1.0 (perfect bouncing) for both the sphere and the ground.
The sphere falls down with a null initial speed. It bounces, and bounces, and bounces again...
But surprisingly, I can clearly observe that the sphere is bouncing higher and higher !
What's going on?
TIA,
PhilB

[PhilB]

None can provide me with an explanation please ? Given the conservation of energy, it should not be possible for a ball to bounce higher and higher, right ?
Thanks
PhilB

[Erwin Coumans]

There is some energy added during collision response, so using 1.0 for restitution is not recommended.

One of the reasons is that collisions usually cause some penetrations, and recovery for those penetrations adds some momemtum (velocity/energy). Usage of split-impulse technique, available in the Bullet 2.72 C++ version, avoids this problem, but introduces some instability for stacked objects.

This is indeed a shortcoming, currently in Bullet and several other physics engines,
Erwin

[PhilB]

Thank you for these explanations Erwin.
Phil