Bullet physics implementation and other stuff

   302819   385   4
User Avatar
Member
134 posts
Joined: March 2009
Offline
Absolutely amazing with this kind of speed! Keep up the good work, i will definitely keep an eye on this thread
Magnus Pettersson
Lead Effects TD @ Storm studios
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
VictorG:
not much:
- completely new layout, where I can create, delete and modify objects in time, as in DOP.
- add some new example scene.
- weariness and refracturing.
- solver output(collision and objects info)
- low/high geometry output
- new source objects for fluids
- visualization of density for fluids


omegan:
thx, speed is very important, because in production We can simulate multiple situation and choose the right


I moved the release of the next version in December , It has few bugs and I have a lot of other work at school.
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
4 posts
Joined: May 2008
Offline
It's awesome cybermax! Will you release a Linux version? and is it possible to go Opensource with this project? I beileve some smart people in houdini community might make it even better.

cybermax
VictorG:
I moved the release of the next version in December , It has few bugs and I have a lot of other work at school.
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
cybermax
update:
ok, I made some changes in btCompoundShape and now its more stable, faster and with any errrors(what I know). It can use different(random) max_force attribute which creating nice destruction effects - for concrete looks better than classic constraint.

what is it changes? did you mean changes in source of native bullet code?
Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
almatea
what is it changes? did you mean changes in source of native bullet code?
NO, btCompoundShape I use for new hard constraint system (something like btFractureBody). I described it wrong, sorry.
However, I changed native bullet source code, due to plugin higher speeds. But these are small changes.

curveU:
linux version will come later, but Its in my plane.
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
cybermax
almatea
what is it changes? did you mean changes in source of native bullet code?
NO, btCompoundShape I use for new hard constraint system (something like btFractureBody). I described it wrong, sorry.

thanks for answer

do you use a constraints for same purpose? (i mean a using constraints for imitation a glue)
do you use only convex bodies or use a concave also (is it Gimpact or compound of convex bodies?)?
Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
almatea
do you use a constraints for same purpose? (i mean a using constraints for imitation a glue)
New “hard” constraint system dont use constraint, Its using btCompoundShape because Its more stable for many “merge” objects than using constraints. But plugin still support spring constraint.

almatea
do you use only convex bodies or use a concave also (is it Gimpact or compound of convex bodies?)?
For concave body I am using my convex decomposition algoritm(so Its using compound convex bodies), because concave bullet body has bad behavior after collision according my research.
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
cybermax
For concave body I am using my convex decomposition algoritm(so Its using compound convex bodies), because concave bullet body has bad behavior after collision according my research.

do your algorithm safe a shape of body? i mean - is it approximate algorithm or not? this algorithm for any body with arbitrary form?

i use representation of shape as set of tetrahedrons.
Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
almatea
do your algorithm safe a shape of body? i mean - is it approximate algorithm or not? this algorithm for any body with arbitrary form?
i use representation of shape as set of tetrahedrons.
Its not approximate algorithm and It dont use tetrahedrons.
And Its run for all shape form.
I dont public more detail, but Try searching some about BSP with PVS. I worked with this algoritms before few years, when I programming my raytrace render, so I made some changes and convex decomposition was done
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
cybermax
I dont public more detail, but Try searching some about BSP with PVS. I worked with this algoritms before few years, when I programming my raytrace render, so I made some changes and convex decomposition was done

thanks, i also make second way for decomposition similar BSP

Attachments:
untitled1.jpg (17.7 KB)

Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
89 posts
Joined: April 2008
Offline
Just curious, what's the issue with gimpact?

The only downside I see is an approx. 3x performance penalty, but I'd think the complexity of decomposition would become prohibitive for large amounts of geo/objects.
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
mrice
Just curious, what's the issue with gimpact?

The only downside I see is an approx. 3x performance penalty, but I'd think the complexity of decomposition would become prohibitive for large amounts of geo/objects.
I dont remember exactly but I think that problem was If I have many object collision with others, like a column formation. Firstly I saw little vibration then reaction forces were increase … in other words, It wasnt stable.

I have similar instability with many constraint, this is reason why I am creating my “hard” constraint system with btCompound object.
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
and it seems that gimpact body can not be a part of btCompound
Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
89 posts
Joined: April 2008
Offline
cybermax
I dont remember exactly but I think that problem was If I have many object collision with others, like a column formation. Firstly I saw little vibration then reaction forces were increase … in other words, It wasnt stable.
.

Could be that gimpact is configured by default to use a SAT algorithm which doesn't appear to work very well. This test [audenmedia.com] has many stacked concave gimpact shapes.

almatea
and it seems that gimpact body can not be a part of btCompound
You'd have to work around that..
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
mrice
Could be that gimpact is configured by default to use a SAT algorithm which doesn't appear to work very well. This test [audenmedia.com] has many stacked concave gimpact shapes.


I did uncomment “#define BULLET_TRIANGLE_COLLISION 1” in btGImpactCollisionAlgorithm.h - then concave bodies have correct collision, but in high stack have penetration after some time even if I use a big fixedTimeStep parameter (about 240 or 480).

And convex bodies as compound shares work mush faster than gImpacts (I mean perfomance of bullet).

Please give more details about your scene - it very interest!
Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
Concave body works fine in my system. After I release next version I will make some test with gimpact again and compare the speed. If gimpact will be quicker, but there will be high stack penetration, I can make option in GUI and users decide what they want


mrice: I see your video and its cool, can you give us more info(sim time, object,constraints count etc…). Do you have problem with many constraints - unstable, object vibration????
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
38 posts
Joined: Oct. 2008
Offline
cybermax
Concave body works fine in my system.

it would be nice to see a video demonstration about this, especially with high stack of gimpacts.
Vladislav Tushevskiy
http://vimeo.com/heliosalmatea/videos [vimeo.com]
User Avatar
Member
255 posts
Joined: Aug. 2009
Offline
almatea
it would be nice to see a video demonstration about this, especially with high stack of gimpacts.
My current system with convex decomposition you can try, because my plugin is free … so feel free If you find problem, let me know.

Gimpacts I will test later, now I implement more important features(concave body works fine)
https://vimeo.com/user3251535 [vimeo.com]
https://twitter.com/milansuk [twitter.com]
https://github.com/milansuk [github.com]
User Avatar
Member
89 posts
Joined: April 2008
Offline
cybermax
mrice: I see your video and its cool, can you give us more info(sim time, object,constraints count etc…). Do you have problem with many constraints - unstable, object vibration????

About 50k explicit objects and 250k constraints. Sim time averaged 2 minutes/frame, but could be alot faster, gimpact collisions were used much more than needed (due to laziness)

I havent seen any instability with constraints. I'm mostly on linux, but the only difference I know of is that bullet on linux doesn't use the Sony vector math library.
User Avatar
Member
3 posts
Joined: July 2006
Offline
My current system with convex decomposition you can try, because my plugin is free …

Is it free as in Open source or free as in just the compiled win64 version?
I would love to compile this for other platforms if possible.
  • Quick Links