Robotics StackExchange | Archived questions

Motion planning with the position target

Hi!

I'm curious about how the move_group_interface realizes the motion planning task with the only position target.

By browsing the source code of movegroupinterface.setPositionTarget, I found that the orientation target is cleared.

In fact, I want to finish a motion planning task with the Python binding of OMPL and don't know how to set only a Cartesian position target instead of a joint value target. For now, I'm trying to find out how MoveIt computes the joint value target with the kinematics plugins.

As far as I know, the kinematics plugin, such as KDL and TRAC-IK, are not collision-aware kinematics algorithms. How do they compute the collision-free configuration which is treated as a goal state for motion planning algorithms?

MoveIt integrates plenty of C++ classes which are difficult for me to find the realization of the motion planning procedure.

Do you have any ideas?

Any help would be appreciated!

Asked by JohnDoe on 2023-07-13 04:20:26 UTC

Comments

Answers