Search found 26 matches

by Johan Gidlund
Thu Mar 29, 2012 12:30 pm
Forum: Career Opportunities
Topic: Frostbite looking for Physics SE
Replies: 0
Views: 32293

Frostbite looking for Physics SE

We are looking for Software Engineers to strengthen the Physics & Destruction group of the Frostbite team. You can find more information at http://dice.se/jobs/sr-engineer-physics/ You can send a mail to johan.gidlundmonten@dice.se if you have any questions about this opening. We also have openi...
by Johan Gidlund
Mon Feb 20, 2012 4:20 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Collision Detection Frontiers/5 Year Projection
Replies: 2
Views: 6113

Re: Collision Detection Frontiers/5 Year Projection

I would say that there's increasing emphasis on using Finite Element Method. It has always been used for structural analysis with varying precision in the old days to be fast enough to calculate. It can be approximately done in real time now, but the problem of optimization still has more room to g...
by Johan Gidlund
Tue Jan 17, 2012 9:50 am
Forum: General Bullet Physics Support and Feedback
Topic: Efficient collision detection for destructible environment
Replies: 3
Views: 9876

Re: Efficient collision detection for destructible environme

I am not familiar with the details of Bullet but from a general point of view you should try to use one rigid body with a list of shapes. Having a rigid body per block is insane and will never work for the amount of pieces you are targeting. For the large amount of pieces you are targeting it is pro...
by Johan Gidlund
Sun Jul 10, 2011 9:57 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: What to study in prep for writing a physics engine?
Replies: 2
Views: 6299

Re: What to study in prep for writing a physics engine?

The math you need to know to write a physics engine has not changed in some time so what was valid in 1997 is probably still valid. What has changed are the actual algorithms that are used for the different stages on the simulation. Christers book should give a pretty good overview of what is typica...
by Johan Gidlund
Tue Mar 22, 2011 8:34 am
Forum: General Bullet Physics Support and Feedback
Topic: SAP or Dynamic tree when simulating LOTS of spheres?
Replies: 2
Views: 3671

Re: SAP or Dynamic tree when simulating LOTS of spheres?

I have not tested this with Bullet but in my SPH fluid simulation I use more than 10k particles with SPH and it runs quite fast. Unless I misunderstand your problem you should also have a situation where the relative position between particles don't change frequently so SAP should be very well suite...
by Johan Gidlund
Mon Jan 31, 2011 1:13 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Bouncing Balls Down Stairs
Replies: 7
Views: 12626

Re: Bouncing Balls Down Stairs

Bounciness is typically controlled by restitution.
by Johan Gidlund
Fri Jan 14, 2011 10:32 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: How to get a "perfect" bounce
Replies: 3
Views: 4966

Re: How to get a "perfect" bounce

Oh you are talking about the actual directions. Restitution has nothing to do with that obviously. What will control the direction is the normal or the collision. Is your ball a sphere shape? If it is a hull or triangulated ball it will not have smooth normals and is thus bound to bounce in a more e...
by Johan Gidlund
Mon Jan 10, 2011 9:18 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: How to get a "perfect" bounce
Replies: 3
Views: 4966

Re: How to get a "perfect" bounce

You should tweak restitution. A value of one should give a fully elastic bounce.
by Johan Gidlund
Thu Dec 16, 2010 5:09 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: The cage
Replies: 3
Views: 4111

Re: The cage

What exactly do you want to happen when objects reach the border of you "cage"? I think I would have used constraints like you first suggested as long as the area objects are allowed to move inside is easily described mathematically. The other object in your constraint would in this case b...
by Johan Gidlund
Wed Dec 15, 2010 11:04 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: How to control forces of kinematic body against rigid body
Replies: 2
Views: 4003

Re: How to control forces of kinematic body against rigid bo

I have not worked with Bullet in a long time so I don't know exactly how things work in it.
But talking for a general perspective mass should effect the force of the collision even for a keyframed object?
Tweaking restitution should also change the behavior.
by Johan Gidlund
Fri Dec 10, 2010 10:04 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Is there anyway around quaternions for setting orientation?
Replies: 2
Views: 4051

Re: Is there anyway around quaternions for setting orientati

If you use matrices for storing transformation you will need to continiously renormalize them to prevent drift.
Read section 4 of this paper for more information.

http://www.cs.cmu.edu/~baraff/pbm/rigid1.pdf
by Johan Gidlund
Fri Dec 10, 2010 9:53 am
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: how to enumerate all contact pairs ?
Replies: 10
Views: 11596

Re: how to enumerate all contact pairs ?

So you propose to make own code to find contact pairs. In this case yes because it's very simple. Checking if two spheres intersect is trivial and checking for two oriented boxes is not hard either. I think there is box-box intersection code in bullet that you could use. Yes, trees introduce some o...
by Johan Gidlund
Thu Dec 09, 2010 6:19 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: GJK on GPU or any other good algorithms?
Replies: 3
Views: 5506

Re: GJK on GPU or any other good algorithms?

If I find time for it I will dig up some paper and post a link. Yes you should do redundant tests. Just remove all the "smart" if cases and do the work. GPU:s can do lots of matrix math but they don't do branching well. Generally speaking you should not do collision detection for complex o...
by Johan Gidlund
Thu Dec 09, 2010 4:45 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: how to enumerate all contact pairs ?
Replies: 10
Views: 11596

Re: how to enumerate all contact pairs ?

If you have a static scene of only boxes common algorithms will probably not be very efficient. Most of them use frame coherency in some way. Why I would do in your case would be to use the bounding sphere of all boxes (because it's trivial to find) and insert them into some kind of tree. From this ...
by Johan Gidlund
Thu Dec 09, 2010 4:32 pm
Forum: Research and development discussion about Collision Detection and Physics Simulation
Topic: Collision Detection Course materials / Slides
Replies: 3
Views: 4525

Re: Collision Detection Course materials / Slides

That material is not beginner level.
Check this page for some material written by Erwin.
http://realtimecollisiondetection.net/pubs/

That book is also good to learn all the needed concepts.