[ros-users] [Discourse.ros.org] [ROS-Industrial] Path planning with external axis
ros.discourse at gmail.com
Mon Sep 10 15:29:24 UTC 2018
I would like to try what @yijiangh mentioned above. In other words, I want to use _descartes_ for redundant robots, where the redundant joints are sampled/discretized. I have plenty of idea's for the sampling part, it is with the actual implementation that I'm stuck. I've looked through the source code of MoveIt! and _descartes_ but I'm not sure what the best approach is.
1. Add functionality to the MoveIt! IK plugin to return a certain number of possible solutions for a redundant robot. There seems to be a lot of functionality present in moveit_core/kinematics_base to work with redundant joints, but I've not figured out how it works yet.
2. Let _descartes_ handle the sampling of the redundant joints. Use separate planning groups for the redundant and non-redundant parts of the robot. This approach seems particularly useful for external turntables.
3. Use a separate package for the inverse kinematics and add redundant joint sampling there. (For example [opw_kinematics](https://github.com/Jmeyer1292/opw_kinematics) Only use MoveIt! for collision detection.
4. Some combination of the above. or other approaches I did not think of.
Approach 1. would make it available to other planners. On the other hand, you could argue that sampling is the task of the motion planner and therefore approach 2. is better. Approach 3. is somewhere in the middle.
I cannot find a lot of information on the topic using google or looking through Github issues.
Can someone help me with pointers / interesting links / insights / discussion?
[Visit Topic](https://discourse.ros.org/t/path-planning-with-external-axis/4451/13) or reply to this email to respond.
More information about the ros-users