Society of Robots - Robot Forum

Mechanics and Construction => Mechanics and Construction => Topic started by: Tantor on March 02, 2009, 06:34:22 PM

Title: 2D Collision test
Post by: Tantor on March 02, 2009, 06:34:22 PM
Hi Everybody!

I'm working on a mechanical dynamics engine. I intend to properly simulate walkers so I'm trying to get right the contact forces agent... I have send on youtube a video of a very simple test showing a 2D rod bouncing on a plane. But I'm not sure if it's right (realistic). contact is frictionless, collisions are tangentially and radially elastic (full rebound)... it is legitimate to wonder if the rod should rebound alternatively from side to side, rather than on the same side, has shown in the video...

        testChoc2D (http://www.youtube.com/watch?v=jd_CTocjZs4#lq-lq2-hq)

thanx you indeed for helping, and please keep an eye on my videos!

Tantor
Title: Re: 2D Collision test
Post by: HDL_CinC_Dragon on March 03, 2009, 03:13:49 PM
Well if you imagine a rod or bar falling like that, the force of its end hitting the plane would enact a force on its entirety which means it would cause the whole bar to bounce upward rather than only the same end that hit. Try getting some pieces of scrap word or metal and drop them in front of a camera (higher FPS the better) and then study the motion it goes through. If you had a standard high speed camera you would be golden :P

So yes the rod should in fact rebound semi-alternately but technically both sides would be bouncing at the same times just not necessarily in the same fashion...

I hope I got my thought across >_<

-Dragon
Title: Re: 2D Collision test
Post by: Tantor on March 04, 2009, 10:38:50 AM
thanx!

right now when the rod is horizontal against the plane, there are two contact points computed (one on each side-end)... perhaps computing more contact points will help...

I just have tried to figure out te right behavious by reproducting the experiment with a simple pen!!! it is difficult to conclude...
Title: Re: 2D Collision test
Post by: HDL_CinC_Dragon on March 04, 2009, 08:20:56 PM
Well heres a question that may or may not help:
What information about the contact points/bar is the program dealing with? Does it just say "This point contacted a solid object, make it move to its original position so it can start over again and fall again" or does it actually say "Object with x mass and y velocity with z elasticity, etc, etc, etc hit an object with x mass and y velocity with z elasticity, etc, etc, etc. in this gravitational(earths gravity - 9.8m/s^2)  environment The resulting force vector would be this much force, this direction, etc etc etc."?

Its a heck of a lot of data to crunch but the more data the more accurate the more useful the more awesome the more money you can get :P
Title: Re: 2D Collision test
Post by: Soeren on March 05, 2009, 02:42:23 AM
Hi

Try getting some pieces of scrap word or metal and drop them in front of a camera (higher FPS the better) and then study the motion it goes through. If you had a standard high speed camera you would be golden :P
Just get the Casio Exilim EX-F1 (or the slightly slower EX-FH20) which goes up to 1'200 fps in film mode, shoots 1920 × 1080 up to 60 fps, macro to 5 cm (2") and a vast range of usefull features.
The cameras web pages start at http://www.exilim.com/intl/ex_f1/ (http://www.exilim.com/intl/ex_f1/)

[youtube]Water Balloon - Casio EXILIM Pro EX-F1 - 600fps (http://www.youtube.com/watch?v=i0SnYaByCek&feature=PlayList&p=A6D776E9EC0DBB51&playnext=1&index=7#lq-lq2-hq)[/youtube]

[youtube]High Speed Slow Motion FRESNEL LENS SUN SOLAR Beer Bottles Casio Exilim Pro EX-F1 (http://www.youtube.com/watch?v=IlnscjPNF8Q&feature=PlayList&p=A6D776E9EC0DBB51&index=9#lq-lq2-hq)[/youtube]

Title: Re: 2D Collision test
Post by: Tantor on March 05, 2009, 06:11:43 PM
The problem is also that it simulates an very idealized, although simple, system. the contact solver uses the signorini conditions... indeed what we see might be a correct solution... difficult indeed...