Search found 16 matches

by irlanrobson
Tue Mar 02, 2021 9:15 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: How can I speed up my SAT test?
Replies: 1
Views: 31272

Re: How can I speed up my SAT test?

You don't need to regenerate all contact points. I either write a separation or overlap cache and check this in the next frame. Then if the shapes are overlapping in the previous and current frame rebuild the contact. In the case of a face overlap check if the relative orientation of the bodies has ...
by irlanrobson
Thu Nov 07, 2019 2:07 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Boucing with sequential impulse
Replies: 12
Views: 57863

Re: Boucing with sequential impulse

Without warm-starting the solver can take lots of iterations to converge. For this reason you should consider using warm-starting. FYI one of the main reasons for a simulation to blow up is a bad choice of integrator. Semi-impicit Euler is stable for non-stiff equations. If your simulation is explod...
by irlanrobson
Mon Nov 04, 2019 4:02 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Boucing with sequential impulse
Replies: 12
Views: 57863

Re: Boucing with sequential impulse

Well, one problem in the code you posted is that it solves the tangent and normal constraints independently. I don't think this is correct. The friction forces depend on the normal forces. Therefore, you should solve a tangent constraint followed by a normal constraint or vice-versa since constraint...
by irlanrobson
Tue Oct 15, 2019 3:42 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Boucing with sequential impulse
Replies: 12
Views: 57863

Re: Boucing with sequential impulse

Your math seems to be correct. There is probably a small problem in your code. Maybe a sign problem or something like this.

Comparing your engine with Box2D may help you to identify the issue:

https://github.com/erincatto/box2d-lite

https://github.com/erincatto/Box2D
by irlanrobson
Tue Jul 30, 2019 5:00 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: State of the art cloth simulation
Replies: 5
Views: 43004

Re: State of the art cloth simulation

It's worth mentioning that there is a book about cloth simulation that was released recently. The author announced the book release in this thread last year: https://pybullet.org/Bullet/phpBB3/viewtopic.php?f=6&t=12356 I don't own one, but it seems a good book, and it has some math. The part abo...
by irlanrobson
Tue Jul 30, 2019 4:53 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: State of the art cloth simulation
Replies: 5
Views: 43004

Re: State of the art cloth simulation

For collision detection between cloth and arbitrary convex/concave meshes you want fast point queries. This can be done using a signed distance field (SDF). For convex shapes, calling GJK thousands of times per step can substantially hit performance. SDF allow O(1) evaluation of distance between a p...
by irlanrobson
Fri Jul 26, 2019 4:03 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: State of the art cloth simulation
Replies: 5
Views: 43004

Re: State of the art cloth simulation

Then there is corotational FEM which can be done realtime. The algorithm complexity of this solver is dominated by the CG method as in David Baraff's cloth solver. You can apply the same constraint satisfaction technique described in Baraff's solver to a corotational FEM solver. For cloth it seems t...
by irlanrobson
Thu Jul 25, 2019 5:47 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: State of the art cloth simulation
Replies: 5
Views: 43004

Re: State of the art cloth simulation

A commonly used approach in the industry is the continuum approach described in the paper of David Baraff. https://www.cs.cmu.edu/~baraff/papers/sig98.pdf However, some forces like the shear (and bend) force can introduce problems into the linear system described above if included incorrectly so it ...
by irlanrobson
Mon Oct 31, 2016 5:56 pm
Forum: General Bullet Physics Support and Feedback
Topic: GJK Algorithm for Quadcopter Collision Detection
Replies: 6
Views: 12468

Re: GJK Algorithm for Quadcopter Collision Detection

In my opinion the GJK presentation by Erin Catto and its source code on the Box2D repository are the best resources available for understanding and implementing the GJK algorithm in 2D. If you need GJK in 3D I recommend the source code I posted. It is basically a 3D version of Box2D's. Here is Erin'...
by irlanrobson
Sun Oct 30, 2016 1:31 am
Forum: General Bullet Physics Support and Feedback
Topic: GJK Algorithm for Quadcopter Collision Detection
Replies: 6
Views: 12468

Re: GJK Algorithm for Quadcopter Collision Detection

wyattsmcall1, I'm not a big fan of throwing my collision modules away but will make an exception here. I attached the GJK submodule that belongs to the collision module of Bounce, a physics engine I've been working on since some months ago. It uses barycentric coordinates and Voronoi regions to find...
by irlanrobson
Wed May 18, 2016 4:06 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Numerical issues with GJK for large triangle areas
Replies: 8
Views: 16817

Re: Numerical issues with GJK for large triangle areas

Yeah. Obviously if the contact point doesn't align with the center of mass of Sphere 1 because of rouding errors being accumulated when projecting its center on the triangle, and if Sphere 2 does, then the latter will start sliding after some significant amount of time has passed.
by irlanrobson
Wed May 18, 2016 1:05 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Numerical issues with GJK for large triangle areas
Replies: 8
Views: 16817

Re: Numerical issues with GJK for large triangle areas

Okay, I tested the stack on a hull and now with the sphere stack starting from a different location, and looks like a numerical problem due to GJK barycentrics. If anyone is able to reproduce this same setup and post back the results I'd greatly appreciate and set this thread as solved. Box (Hull) P...
by irlanrobson
Tue May 17, 2016 9:22 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Numerical issues with GJK for large triangle areas
Replies: 8
Views: 16817

Re: Numerical issues with GJK for large triangle areas

The support point is just the sphere center. I'm pretty sure the implementation is correct as it is has been working very well since implemented. I just need to know if this is a common numerical problem for GJK (large shape sizes). I've read about this in the forums some weeks ago but can't remembe...
by irlanrobson
Mon May 16, 2016 12:58 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Numerical issues with GJK for large triangle areas
Replies: 8
Views: 16817

Numerical issues with GJK for large triangle areas

I've noticed that my GJK implementation is suffering from numerical issues for a 3-simplex early out (e.g. the closest point is on a triangle), where the shapes involved are a sphere and a triangle. What made me notice this was a vertical stack of 10 spheres falling over the following static triangl...