On Sun, 28 Dec 2008, manuelbirlo [..] ... wrote:

> Is the ChainJntToJacSolver::JntToJac() a function with optimized runtime
> efficiency? Is is the best way to calculate a jacobian?
>
> I've written my own function to create a jacobian in a geometric manner
> via cross products. It produces the same results. And now, I'm trying to
> find a reason to use it instead of JntToJac :-)

If you can exploit specific geometric properties of your kinematic chain,
the general purpose Jacobian computation will always loose... However,
that's why we have the concept of "kinematic families" in KDL, together
with the possibility to provide different "solvers" for the same interface.
Hence, if your kinematic chain indeed has specific properties, it is
worthwhile to create a new kinematic family, identifying its specific
properties, and providing a customized solver for it.

Herman

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm