Re: [eigen] geometry module

[Mathieu Gautier <mathieu.gautier@xxxxxx>]

Hi,

> That said, either what you propose (quaternion+vector) or a "dual
> quaternion" class as has been discussed previously on this list, would
> be interesting. We just happen not to have that yet.

I have indeed proposed a representation of 3D position, velocity and 
forces based on the lie group formalism. (see link at the end to the 
mail). With position stored as quaternion and vector3, both aligned and 
3d velocity and force inherited from Matrix<Scalar, 6 ,1>. This module 
could easily be extended to a 2D setup since, a 2D position could also 
be represented as an element of a Lie Group. I also have some functions 
implemented such as log and exponential used to derivate or integrate 
positions and velocities.

I recently modify my code to use a recent revision of the Eigen dev 
branch. But, there are many missing things, documentation, unit tests, 
alignment for map object and interface with Transform objects. I also 
had some inlining issues with visual studio. I hadn't find enough spared 
time to do this, that's why I couldn't provide anything since last april.

By the way, I found that I didn't provide the few slides I made for the 
last meeting. I put them in the attachement.

quaternion + vector : lie group formalism
http://listengine.tuxfamily.org/lists.tuxfamily.org/eigen/2009/11/msg00211.html

double quaternion :
http://listengine.tuxfamily.org/lists.tuxfamily.org/eigen/2009/09/msg00098.html


--
Mathieu Gautier

Attachment: eigen_lie_pres.pdf
Description: Adobe PDF document