Poor MoveIt Paths, caused by IK solver or path planner?

asked 2021-01-18 02:17:46 -0500

iwuur gravatar image

Hi,

I am looking to use MoveIt with my UR10e robot. To integrate well I use this (https://github.com/ros-industrial/uni...) package. My ROS distro is Noetic. In general everything works fine but MoveIt sometimes comes up with crazy paths and sometimes doesn't find a path at all.

My question regards the way moveit plans a path. So lets say I want to move my end effector from one point in cartesian space to another. Then how does MoveIt proceed? My understanding is that first the IK solution is computed for the goal point and then the path planning kicks in and plans from the current joint configuration to the goal configuration without any collisions. Is that correct?

Because the goal joint configurations are often near the joint limits, I get lots of problems with paths after that. This seems to me to be a problem with the IK solver and not the path planner, because a path is found to the new joint configuration but the joint configuration is just not good.

How would I go about solving this problem? I guess I would have to try different IK solvers but I have already done that. I tried the IK solver provided in the above mentioned package, and also the two solver mentioned here (https://ros-planning.github.io/moveit...). This did not improve my results however.

Any comments are appreciated. Also if some of my assumptions are wrong I am happy to be corrected.

Thanks

Edit: Actually the IK-Solver included in the above mentioned package shows solutions when moving the 6 DOF marker in Rviz. But when I plan using the moveit cpp API it does not find solutions. That is very weird to me.

edit retag flag offensive close merge delete

Comments

Did you solve this issue, I'm facing the same

Arn gravatar image Arn  ( 2021-12-25 05:55:40 -0500 )edit