Gyroscopes?

[coder0xff]

I read another (rather old) thread that suggests there is some level of gyroscopic effects, but that Coriolis effect wasn't implemented (which I don't see how that's related really, and should be a natural result of momentum).

So considering the current version, how accurately/stably are gyroscopes simulated?

[Dirk Gregorius]

The Newton-Euler euqations have the following form [1]:

M*v' = F
I*w' = tau - w x I * w

The second term of the angular equation is usually dropped for stability reasons in game physic engines such as e.g. Bullet. Of course you can apply these forces yourself each frame using the applyTorque() interface. You just have to be careful if you take several substeps since this force needs to be re-evaluated for each substep.


Cheers,
-Dirk


[1]
http://en.wikipedia.org/wiki/Newton%E2% ... _equations

[coder0xff]

I'm no mathematics or physics expert, so I'm likely wrong, but I don't see how the second term of the second equations models gyroscopes. Best way I can explain is by example. In the case of a sphere the second term would evaluate to a zero length vector (the cross product of two co-linear vectors), but a sphere can still have precession. I use a sphere as example because its inertial tensor doesn't change the direction of the vector.

As far as I can tell... the second term accounts for changes in angular velocity due to the rotation of the inertial tensor, since angular velocity can change without applying any torque. However, the definition of tao in the wikipedia article says it's total torque, which I suspect is supposed to include gyroscope effects. I'm guessing that the second term isn't included in integration (but as discreet steps) since it would form a differential equation more difficult to solve than it's worth. Updating the orientation of the rotational inertia each frame is more than adequate for plausible simulation.