I've written a function which should calculate the manipulability measure of my KDL::Chain. For this purpose I create a JntToJacSolver an execute the JntToJac-function with the current joint values as input.
So I get the resulting jacobian for the manipulability measure. Now my question is: leads this way to create a Jacobian to a correct manipulability measure sqrt(det(JJ^t))?
I have always values between 15000 and 20000 during a trajectory. Is this correct?
By the way: dou you have some papers about this topic?
manipulability measure calculation with KDL::Jacobian
On Mon, 29 Sep 2008, manuelbirlo [..] ... wrote:
> I've written a function which should calculate the manipulability measure of
> my KDL::Chain. For this purpose I create a JntToJacSolver an execute the
> JntToJac-function with the current joint values as input. So I get the
> resulting jacobian for the manipulability measure. Now my question is: leads
> this way to create a Jacobian to a correct manipulability measure
> sqrt(det(JJ^t))?
If you do the calculation for the matrix multiplication, the determinant
and the sqrt correctly, yes! :-)
> I have always values between 15000 and 20000 during a trajectory. Is this
> correct?
No idea! This depends on so many things: chosen units, chosen reference
frames, actual configuration, ...
> By the way: dou you have some papers about this topic?
For example:
@Article{ doty-etal-itra95,
author = {Doty, Keith L. and Melchiorri, Claudio and Schwartz,
Eric
M. and Bonivento, Claudio},
title = {Robot manipulability},
journal = {IEEE Trans. Robotics and Automation},
year = {1995},
volume = {11},
number = {3},
pages = {462--468}
}
Herman
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
manipulability measure calculation with KDL::Jacobian
I've found this paper (Robot manipulability) but I have no access to view it:
http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel4/70/8824/00388791.pdf?tp=&isnumber=&arnumber=388791
or do you have it?
>If you do the calculation for the matrix multiplication, the >determinant
>and the sqrt correctly, yes! :-)
then it should be correct!
>No idea! This depends on so many things: chosen units, >chosen reference
>frames, actual configuration, ...
hmm...I have a KDL::Chain with 7 joints and an extra none-segment which is for the transformation in woorld coordinates, so the Chain has 8 segments (first segment is a none segment, and 7 segments with 7 rotational joints).
and the joints values are in radians.