Do free-rotating objects always revert to axis with minimal inertia?
Yeah.. i know, question was a mouthful. I just threw together a rigid body simulator and I am rotating boxes of various dimensions. What I find is that no matter what angular momentum I impart, the box (after a little instability) ends up rotating with its long axis along the axis of the angular momentum. Now this sounds reasonable, but for my last simulator a body would constantly rotate about weird random axes, never achieving a semblance of stability. Which way is right.. Assuming no torque and only an initial angular momentum, does a box always end up spinning relatively stably with its long axis parallel to the angular momentum axis, or should it be more unstable?
Thanks in advance,
Terrence
No, sounds like you implemented the good old and incorrect RBD.
None of the methods I''ve seen so far seem to work correctly, especially when it comes to nutation. The "stabilization" you observe is an artifact of analytical and numerical errors.
Take a look at this mod I made to Chris Hecker''s physics demo:
http://www.geocities.com/bpj1138/gdphys3d_test.zip
As you can see the object quickly stops nutating and just spins around the long axis, much like what you''re talking about. What should happen is the object should nutate indefinitely.
Now, this is a controversial topic. Questioning 17''s century mathematics is never easy. There are some people trying to improve the accuracy like Samuel Buss, there is even a new form of math called Geometric Algebra that seems to deal with this topic. So in the words of some usenet poster, "you''ve just entered the biggest headache known to math, good luck."
None of the methods I''ve seen so far seem to work correctly, especially when it comes to nutation. The "stabilization" you observe is an artifact of analytical and numerical errors.
Take a look at this mod I made to Chris Hecker''s physics demo:
http://www.geocities.com/bpj1138/gdphys3d_test.zip
As you can see the object quickly stops nutating and just spins around the long axis, much like what you''re talking about. What should happen is the object should nutate indefinitely.
Now, this is a controversial topic. Questioning 17''s century mathematics is never easy. There are some people trying to improve the accuracy like Samuel Buss, there is even a new form of math called Geometric Algebra that seems to deal with this topic. So in the words of some usenet poster, "you''ve just entered the biggest headache known to math, good luck."
Bingo! That demo couldn''t have captured it better. Well thanks for the heads up.. if you''re still around, what do you suggest to approximate the motion reasonably? The way I see it I either change the integration scheme (I''m implementing Runge-Kutta right now), lower the step size, or actually physically alterthe equations of motion (probably not a good idea). Are those really the only practical things that can be done to get as accurate a simulator as possible?
What code to you want exactly? unlike you i dont have a site to dump it to.. as far as the effect goes its almost identical to that of the modified demo you showed, it basically moves from its initial position to the position where it rotates about its longest axis. Strangely enough I made a DIFFERENT simulator a while ago ( i was supposed to do it right this time, ha ) and objects in that seems to nutate just fine.
Hrm, there is lots of free sites around, I use geocities myself. Why don''t you make the same test with the code you had before and post it (at least the binary).
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement