It is nearly impossible to, in real life, spin a pencil fast enough to keep it from falling over. But this simulated cylinder does stay up for a while with only a gentile spin. YES I remembered to do alt+r and ctrl+a after scaling the object. I know it’s a LOT bigger than a pencil, but even an object this size shouldn’t behave this way.
Thanks. Just like other physics engines, Bullet favors stability over realism. It might be due to lack of gyroscopic terms.
There is no reason to ‘spread the word’ or ‘shout it off the roofs’: I’m sure we all can easily find shortcomings and ‘unrealistic’ behaviour. The main challenge in using rigid body dynamics at the moment is making use of it, while keeping its limitations in mind. I’ll put it in the ‘Physics Tips’ here:http://www.continuousphysics.com/mediawiki-1.5.8/index.php?title=Physics_Tips
I will check it out at some stage, and see if there are some improvements possible.
Thanks for the feedback,
Erwin
This is not a example of bad physics but typical issues in computers. In blender the pencile is perfectly balanced. There are no random forces from air flow or a rough surface. There is also no varaion in mass throught the pencile. The only “randomness” is from rounding errors.
All physic simulations are just that, simulations. They all in one way or another approximate the real world physics, but also capture some real elements of the underling physical process.
Also, if 1 blender unit is equal to 1 meter, a pencil that is maybe 10 centimeters long in Blender would behave more realistically. But, since there is a small invisible shell around smaller objects, this won’t work. Until the bounding fits the object exactly, you will have to make things bigger.
Show me a physics engine that likes Long narrow objects.
I have tried havoc, Newton and a few others in commercial games
They all have problems/limitations with this sort of thing.
Part of game design is taking what resources you have and making something nice with it.
That is all part of the sport…
Whatever you cant do, you fake
I am very happy with the current bullet, and I have toyed with the next version as well. Things are only going to get better.
What are you talking about. Long narrow objects pose no particular problem to any physics engine that I know of. I has a stable moment of intertia so the eular equations are easy to solve and are stable. Groups of object with collsions is a different matter entirly. In fact its still a resurch problem.
However i totaly aggree you need to work with the limitations of the engine you have. They all have them and Bullet is a pretty good RT physics engine, so will lilkly have the same issues in other engines.
Also remeber that what you think is realistic may not be so. In particular objects with rotation do not behave intuativly. aka gyroscopes and the like. There is some bizzare and down right unpredictable things that can happen…
I updated the demo with a workaround (compound objects). I am astonished by how stable and realistic this workaround is, but it is slow (not the demo, but it would be bad with 10 or so).
I’m not saying that bullet is bad, just that it has a problem that should be fixed. Otherwise bullet is amazing.
Performance will always be an issue. Well untill everone has a 100 teraflop machine in there pocket. Some of the physics is hard and slow, there is no way around it. (and the physics card by some accounts makes it slower). You have to work with these limitations to get the Ohhhh factor still going. Thats what its all about.