Rigid physics bug

[0xdeadbeef]

When crashing, G4 sometimes performs some pretty wild physics calculations. I found this especially easy to reproduce with the Wildcatfish (http://www.knifeedge.com/forums/showpost.php?p=98999&postcount=45). I appended a recording at Alpine Lake with this nice plane by Dhk79&Cowboy.

Note that this isn't even the worst thing I've seen yet. I had situations were the fuselage would spin around itself in midair for a minute or so. Not only are the collision forces obviously way off, this also seems to defy gravity.

For the fun of it, I also made a video of the recording (15MB):
http://rapidshare.com/files/75001093/G4_Physics_Bug.avi.html

[jbourke]

Do you see this problem with stock aircraft or custom aircraft or both?

This seems to happen a lot with custom aircraft because the designers don't close the collision mesh correctly. I'm not sure if that is the case here. I'm just speculating.

Jim

[pplace]

don't close the collision mesh correctly

Just wondering what you mean by "close"

[0xdeadbeef]

This happens with stock planes as well, but less spectacular and less reproducible.
Anyway, after only a few trials, I got a quite nice one with a "Big Stik Float Plane" on Alpine Lake. See attachment.

[nemo_uk]

Just wondering what you mean by "close"

An easy way to check if there are any problems if you are using 3ds max is to apply the STL check modifier to the piece of geometry (collision object) you are checking.

[pplace]

thanks nemo,

I didn't know if it was more of a "technical" term he was describing. Turns out it wasn't.

I was actually thinking along the lines of not adding collision meshes to every part of the aircraft (as they suggest in the tutorials) I have never done this....For example I would add collision meshes to the wheels, but not necessary the landing gear always. (As it is nearly impossible to bump, or hit the landing gear without hitting the wheel first)

So I wanted to make sure "closing" didn't mean in a round about way.....leaving objects un collision meshed.

Thanks

[ladams]

Actually I've seen this all over the place, and not just with the airplanes.

The physics dose not clamp properly and TINY variations in interactions between dynamic parts can cause things to go flying wildly. For exmaple a wing piece settling on the ground after a crash will suddenly FLY back up in to the air because of some tiny bump in the ground or something.

Another example, try placing some of those "message in a bottle" objects on top of a wall or another piece of wood in the editor. Or try laying a piece of wood diagonally against a low wall. They just will NOT stay. They start bouncing and jiggling and eventually fall over sometimes hitting the ground and then spinning on axis for a minute or two.

Though RealFlight is not the only program with this problem. Almost all physics engines have this problem, though I have to admit that in real flight it's not the best I've seen...

[0xdeadbeef]

It's a common problem of physics engines that objects don't find a stable rest position. E.g. in Crysis I just had the problem that somehow a gun was stuck in the wall and kept jiggling around.

Then again, the rigid bodies engine in G4 shows two behaviors that other engines don't show: firstly the kinetic energy seems to increase suddenly which catapults parts in the sky or to the horizon after only subtle rebounds. Secondly, parts are rotating in midair "forever" which obviously means that either gravitational force doesn't really have effect on them or they create lift somehow. Both is weird to say the least.