By: nobody ( Nobody/Anonymous )
Quaternion Inversion
20030419 00:14
Reading your page on quaternions described as an angleaxis representation where
the axis consisted of complex numbers has helped me out. I was faced with a
situation where I had to invert a rotation. Previously I had been converting
the quaternion to a rotation matrix then finding the Inversion to the rotation
matrix and converting the inverted matrix back into a quaternion. After reading
your page I tried negating the w (angle value) in the quaternion. This is just
rotating about the same axis as the original quaternion in the opposite direction.
It seems to work, however I would appreciate any comments if someone knows of
something wrong with doing this. Anyways this is much faster than inverting
the matrix.
By: martinbaker ( Martin Baker )
RE: Quaternion Inversion
20030419 07:03
The inverse of a quaternion number is its conjugate (assuming that it is normalised,
which it will be if it is representing a rotation) .
conj(a + b i + c j + d k) = a  b i  c j  d k
So instead of taking minus the real value, as you did, try taking minus all the complex values.
Looking at the web page,
http://www.euclideanspace.com/maths/algebra/realNormedAlgebra/quaternions/index.htm
this was not explained very well, so Ill try to improve it. It helps to hear
from you so that I can get an idea of what works.
Cheers,
Martin
By: nobody ( Nobody/Anonymous )
RE: Quaternion Inversion
20030419 17:57
I have tried what you have suggested and it works also. Both methods appear
to work identically for me although I suspect there may be weird cases where
my method fails. Intuitively it seems to me inverting the axis will result in
the same rotation as inverting the angle. After all you are rotating about the
same axis ( if you extend the original axis through the center point you will
be on the inverted axis ) just in the opposite direction as the original rotation.
There may be some weird symmetry case which I have not considered. There should
be a proof somwhere saying these are or are not equivalent rotations which would
be interesting to see. I think that this could apply to regular angle axis representation
of angles also. I am using your method because I believe you have more of a
mathmatical body of work supporting you.
To get from a quaternion with minus the imaginary parts to a matrix to a quaternion with minus the real part we have to take minus all values.
It seems to me that (qw,qx,qy,qz) and (qw,qx,qy,qz) will produce exactly the same rotation because if we look at the quaternion to matrix conversion all the terms consist of a product of 2 of these elements so the minus values will cancel out.
Thanks for pointing this out,
Martin
metadata block 

see also:  
Correspondence about this page  
Book Shop  Further reading. Where I can, I have put links to Amazon for books that are relevant to the subject, click on the appropriate country flag to get more details of the book or to buy it from them. 

Terminology and Notation Specific to this page here: 

This site may have errors. Don't use for critical systems.
Copyright (c) 19982016 Martin John Baker  All rights reserved  privacy policy.