[ODE] Rotation Matrix Error Correction

[Manohar B.S]

Hello,

 Some time back I came across the problem of 'finite
register length effect' in rotation matrices. It
happens when we use rotation matix recursively.

 Finally I could get a solution to bring Rot matrix
back to track using the following eqn. I donno if it's
worse than what Martin CM told.

 N = M / sqrt(Mt*M).............. eqn1

 N is the least-sq closest rot matrix without errors.
 M is the rot mat with errors. (non-orthonormal)
 Mt is transpose of M.

 Square root is found from cholesky demoposition.
 Mt * M forms a symmetric and +ve defnite.
 Please refer 'Numerical recipes in c' book for the.
 algorithm
 
 eqn1 make gives N which is orthonormal.

 Hope you ppl find it useful.

 --Manohar
 

  






--- "Martin C. Martin" <martin at metahuman.org> wrote:
>  > Is there any ODE function to normalize a rotation
> matrix?
> 
> Convert it to a quaternion and back again.  Even
> better, do all your 
> rotation math with quaternions.  I think that's what
> ODE does internally.
> 
>  > In general it would be nice with an overview and
> maybe some quick
>  > documentation over the math functions available
> in ODE
> 
> That would be great.  In the mean time, try browsing
> the header files, 
> e.g. rotation.h
> 
> - Martin
> 
> 
> 
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode


__________________________________
Do you Yahoo!?
New Yahoo! Photos - easier uploading and sharing.
http://photos.yahoo.com/