[MoveIt!] Goal tolerance for inverse kinematics

asked 2017-07-31 08:47:23 -0500

F4bich
79 ●7 ●10 ●14

Hi there!

I implemented a small program to test for a sequence of positions if a solution for inverse kinematics exists. I used some of the code from the tutorial. Now, I would like to set the GoalOrientationTolerance and GoalPositionTolerance as it can be done in the move_group API, however it seems like I only can set the number of attempts and the timeout.

Is there a way to initiate inverse kinematics computation with different tolerances? And what is the default tolerance?

My thanks, F4bich

edit retag flag offensive close merge delete

add a comment

1

answered 2017-08-02 06:52:42 -0500

v4hn
2950 ●18 ●53 ●57

updated 2017-08-02 06:56:55 -0500

Hey F4bich,

good to hear you are still working with this! :)

Sadly, there is no exposed interface at the moment to set the tolerance per IK request. I'm thinking about more generic interfaces for IK plugins at the moment, but nothing has been implemented yet.

The general default tolerance for IK plugins in MoveIt is 1e-5, although every plugin specifies this on their own and it also remains for the plugin to decide what that tolerance means w.r.t. the combination of position & orientation differences.

The kdl plugin reads a ros parameter "epsilon" that allows to set the tolerance during startup. TracIK does not export its epsilon as a parameter, but I'm sure they would welcome a pull-request to change that though.

As jrgnicho already proposed you could use the low-level functions of KDL/TracIK where you can pass in tolerance as a parameter.

-- Addition --

Please be aware that there is a fundamental difference between Goal*Tolerance for MoveGroup requests and epsilon/tolerance values for IK requests. MoveIt samples from the regions specified by the former and passes the sampled poses to the IK plugin to generate goal poses for planning. The latter are directly passed to the IK plugin.