Question about damping

[BLK Dragon]

According to docs/code, damping in Bullet is value in range 0-1 which means fraction of velocity lost per second.

Since value of damping can't be higher that 1, that means it takes at least one second for rigid-body to stop (with no any forces/torques applied).

But what if I need to stop rigid-body faster?

[Dr.Shepherd]

apply the force/torque backwards? or more straightforward setLinearVelocity/setAngularVelocity ?

This is the simplest way I can come up with.

Cheers

[BLK Dragon]

apply the force/torque backwards? or more straightforward setLinearVelocity/setAngularVelocity ?

Well, applying additional force/torque may cause movement in opposite direction, and tweaking that will be a bit complicated (compared to tweaking single number of damping).

What I really need is ability to specify damping as "time amount needed for rigid-body to stop motion completely (in absence of applied forces/torques)", it's more intuitive IMO. Looks like I just have to implement such version of damping.

[dphil]

If you don't take an object's current velocity into account, how would you produce an intuitive result? Ie stating that an object should come to rest after 1 second, regardless of whether it's traveling 0.1m/s or 10000m/s, doesn't seem like it would give an intuitive result. But obviously if you do need something like this for your own purposes anyway, then you can of course implement it.

[Dr.Shepherd]

Yeah, setting the velocity directly to 0 would result in a sudden stop. I don't know if there is existing methods to stop it continuously to zero.

Force and torque is a bit tricky, I guess, perhaps we should develop some function to do this continuous stop.

[BLK Dragon]

If you don't take an object's current velocity into account, how would you produce an intuitive result? Ie stating that an object should come to rest after 1 second, regardless of whether it's traveling 0.1m/s or 10000m/s, doesn't seem like it would give an intuitive result. But obviously if you do need something like this for your own purposes anyway, then you can of course implement it.

Well, I won't argue that my 'definition' of damping is more intuitive (or it's even need to work that way).
I just need (in certain cases) linear/angular velocity to be damped 'harder' than it does with damping set to 0.999 in current Bullet implementation.

for example doing

Code: Select all

    
m_angularVelocity *= k * btPow(btScalar(1)-m_angularDamping, timeStep);

in btRigidBody::applyDamping (with k hardcoded to 0.2 for quick-test) gives exactly what I want -- with large enough torque movement starts fast and stops fast;
Perhaps I should just add some 'extra-damping-scale' parameter to my rigid-body and set it to 1.0f by default.

[Dr.Shepherd]

If you don't take an object's current velocity into account, how would you produce an intuitive result? Ie stating that an object should come to rest after 1 second, regardless of whether it's traveling 0.1m/s or 10000m/s, doesn't seem like it would give an intuitive result. But obviously if you do need something like this for your own purposes anyway, then you can of course implement it.

Well, I won't argue that my 'definition' of damping is more intuitive (or it's even need to work that way).
I just need (in certain cases) linear/angular velocity to be damped 'harder' than it does with damping set to 0.999 in current Bullet implementation.

for example doing

Code: Select all

    
m_angularVelocity *= k * btPow(btScalar(1)-m_angularDamping, timeStep);

in btRigidBody::applyDamping (with k hardcoded to 0.2 for quick-test) gives exactly what I want -- with large enough torque movement starts fast and stops fast;
Perhaps I should just add some 'extra-damping-scale' parameter to my rigid-body and set it to 1.0f by default.

Yes, that's a simple solution to re-implement the damping method. Straightforward and efficient !