How to measure penetration depth / serialize bugs

Report any bugs here and we'll post fixes

Moderators: Sascha Willems, Thomas

How to measure penetration depth / serialize bugs

Postby JoeJ » Sat Dec 22, 2012 6:25 pm

I have a little problem. At the screenshoot you see ragdoll feet and that there is more penetration at the feets back (where the mass impact is big), than near the toes. Is there a way to measure that penetration per contact? I'd need to feed that information to ai, so it can compensate that.


And the other question: I still see problems related to the capsule - picking joint - collision problem a while ago.
I observe following issues, but i'm not totaly sure with all of them:
* Penetration gets deeper with time and after a minute that brings the ragdoll to fall - with old collision system it stood still forever.
* Iv'e tried spheres, capsules, boxes - but it's the same for all of them
* I get a short jitter attack at periodic time intervals, like this: Each second the contact forces jitter wildely for a short moment and get smooth again afterwards - maybe this is where penetration gets deeper.
* Using capsules, the contact point jumps around (It seems sometimes it's on the surface of the ground, and sometimes it's on the surface of the capsule)
* I saw a single capsule falling to floor (capsule long axis perpendicular to floor normal), but instead of jumping up after collision, it jumped sideways.
* I get asserts (sorry, i need to note them in the future... i'm note in a hurry with that issues)

The question is: If i try to build a test scene with joints from joint lib to reproduce this stuff, will the joints get serialized too?
Attachments
contact.JPG
contact.JPG (28.24 KiB) Viewed 4741 times
User avatar
JoeJ
 
Posts: 1453
Joined: Tue Dec 21, 2010 6:18 pm

Re: How to measure penetration depth / serialize bugs

Postby JoeJ » Sun Dec 23, 2012 1:03 pm

Stupid me, i can get the wanted error correction by building a rotation from foot up vector to contact normals - so no more need to measure penetration :)

I updated newton some days ago and the asserts are very rare now (no one today), so some of the points listed above may be gone as well.
But the sinking into the floor and jittering still happens.
I plan to build a test scene where a box with kinematic joint resting on smaller objects pushes them into the floor.
Just let me know if the joint will be serialized.
User avatar
JoeJ
 
Posts: 1453
Joined: Tue Dec 21, 2010 6:18 pm

Re: How to measure penetration depth / serialize bugs

Postby Julio Jerez » Sun Dec 23, 2012 2:27 pm

do you have a test demo that I can try?
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: How to measure penetration depth / serialize bugs

Postby JoeJ » Sun Dec 23, 2012 2:52 pm

Yes, i switched again to compile newton along wiyh my own code - is that ok or do i need to change to dll?
Release or Debug build?
For the feet i actually use cubes instead of spheres - would you like spheres more (less contacts)?
User avatar
JoeJ
 
Posts: 1453
Joined: Tue Dec 21, 2010 6:18 pm

Re: How to measure penetration depth / serialize bugs

Postby Julio Jerez » Sun Dec 23, 2012 4:04 pm

link to the newton dll, in release
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: How to measure penetration depth / serialize bugs

Postby JoeJ » Sun Dec 23, 2012 5:06 pm

Ok, release mode does not work at the moment and i need to finish the stuff currently in my mind first.
Will become past christmas...
User avatar
JoeJ
 
Posts: 1453
Joined: Tue Dec 21, 2010 6:18 pm


Return to Bugs and Fixes

Who is online

Users browsing this forum: No registered users and 6 guests

cron