[ODE] RE: Angular velocity

[Yefei He]

Jon, 

    After I looked at the code for dRFromEulerAngles(), I took 
[_R(0,0),  _R(0,1),  _R(0,2)] as the x axis of the body frame 
coordinate system after rotation, [_R(1,0),  _R(1,1),  _R(1,2)] 
as the y axis, and [_R(2,0),  _R(2,1),  _R(2,2)] as the z 
axis. In this way, results look correct if I just negate all 
the angular velocity input and output. If I'm to use 
[_R(0,0),  _R(1,0),  _R(2,0)] as local x axi, etc., then I 
should be able to use angular velocity as is. But then there 
will be no way I can set the Euler angles properly, no matter 
what permutation of phi, theta and psi and positive/negative 
values. I'll have to change the ODE code to transpose the 
resulting matrix of dRFromEulerAngles(). 

    Thanks for taking your time to discuss the issue with me. At 
least I know I'm not the only one who found the inconsistency. 

    Best Regards, 
  
    Yefei 


> -----Original Message-----
> From: Jon Watte [mailto:hplus-ode at mindcontrol.org] 
> Sent: Thursday, April 14, 2005 6:24 PM
> To: Yefei He
> Cc: ode at q12.org
> Subject: RE: Angular velocity
> 
> 
> 
> > I interprete _R(m, n) as mth row and nth column, so my 
> layout of the matrix 
> >    _R(0,0)  _R(0,1)  _R(0,2) 
> >    _R(1,0)  _R(1,1)  _R(1,2) 
> >    _R(2,0)  _R(2,1)  _R(2,2)
> 
> That still doesn't say anything, because you need to be 
> specific about 
> whether you are using row vectors (on the left) or column vectors (on 
> the right), AND you need to specify whether the layout is row 
> major or 
> column major.
> 
> IIRC, OpenGL is row major, assuming row vectors on the left, which 
> means the same thing as being column major, assuming column 
> vectors on 
> the right. I believe ODE actually flips one of these conventions -- 
> which of the two you choose to flip really doesn't matter, as 
> the data 
> comes out the same anyway.
> 
> >     and so matrix multiplication is applied in the left to right 
> > order. If
> > that's the case, then the function dRFromAxisAngle() is 
> wrong, and angular 
> > velocity is wrong too, because ODE has: 
> 
> I'm pretty sure angular velocity is right, so instead the matrix is 
> not transposed, but instead the storage and/or vector convention is 
> different from your expectation. Which means that likely the 
> Euler angle 
> functions are the wrong ones.
> 
> Cheers,
> 
> 			/ h+
>