Universal joint ready for testing in svn

[mo]

Hi

There is a new UniversalJoint implementation ready in the svn. I did
som preliminary interactive tests, and they seemed to work out fine,
but there could still be bugs. Let me know how it works out.

[Mikkel Petersen]

That sounds great ! I'm a little busy with work these days, but will
try it out soon as possible, and send feedback here right away-

[Mikkel Petersen]

Seems to work, however with some joints there are some problems..they
start to rotate violently, even when not using motors.
It's hingejoints though. I'll upload an applet soon as possible so you
can see for yourself.

[mo]

Okay. I've encountered violent movements with joints if the limits are
too close to PI or -PI. What limits do you have on your hinge joints?

[Mikkel Petersen]

The limits are not near to PI, also if set the limits to 0.1 or
something, it the same. Its only with two joints, the elbows, and its
a bit strange, because knees are also jointhinges, and there is no
problem with them. I'll post the applet tomorrow.

[Mikkel Petersen]

Ok finally got an applet ready which shows the problem. The arms, or
elbows, are the problems.
This is strange, because the legs (knees) are created exact same way.

Here is the link:

http://onlineboxing.net/ob3d/applet/jpctapplet/test.html

[mo]

Nice little applet by the way :)

Sure that looks strange. Apart from the elbows, the rag doll seems
sound.

Are you sure that inter-doll-body collisions are all turned off? To be
sure, you could disable contact constraints completely, for all
possible pairs. I don't think that it is the issue though...

btw, I noticed that the box definition in jinngine is not given by
half-extends, but by full side lengths of the box. Could that
influence your code in any way?

It looks to me as though some force is pushing the joint against its
will pretty violently. So i guess there are 3 things that could be
happening:

1. There is a bug in the UniversalJoint and/or HingeJoint code
2. There are conflicting joints working, like some bodies are
connected by extra joints by mistake.
3. Some contact constraint are working (which it should not) due to
colliding physics geometry

I think it would be really nice to have a way of exporting the
jinngine setup to a file, such that it can be reproduced easily. I'll
think of a way to do that.

In the mean time, I will try to create a rag-doll setup, much like
yours, and see if I run into something similar.

[Mikkel Petersen]

So I should maybe * x,y,z length with 2 ? I just did and it made a
little difference but they are still acting weird.

I'll keep hunting..

[mo]

Hi again

I committed RagdollExample to the examples project.

http://code.google.com/p/jinngine/source/browse/trunk/jinngine.examples/src/jinngine/examples/RagdollExample.java

Try and update the jinngine.examples project, and run the
RagdollExample. It shows a ragdoll that floats around, much like your
demo, except the rag doll is somewhat simpler, and hardcoded by me, so
its a bit ugly actually :) You can drag around the doll using the
mouse. Hold down space to manipulate vertically. Using this example, I
have been unable to reproduce the weird error we see in your applet.
It is possible to create some jitter by moving the doll around in
certain ways though. I know this doesn't solve your problem, but it
may give us an idea about where to look for your bug...

btw, are you sure that all the bodies that are used in the ragdoll is
actually added to jinngine using Scene.addBody()? When I did my
little ragdoll, I actually forgot that for some bodies by mistake, and
it produced some weird results, at little like what I see in your
applet.

[Mikkel Petersen]

Yes your ragdoll works well..I'll take a look on how you've
implemented it, thanks !

On Jul 23, 11:24 am, mo <mor...@silcowitz.dk> wrote:
> Hi again
>
> I committed RagdollExample to the examples project.
>

> http://code.google.com/p/jinngine/source/browse/trunk/jinngine.exampl...