OMPL planner changes the position of object after planning.
Posted: 17 Jan 2019, 17:10
This post is related to my other post "Changing parent of object changes its position (by a small numerical error)". I decided to create this post, because the other post is mainly about the issue that changing the parent affects the position of the object. The other post is also much more detailed. This post is more about the OMPL planner (might) affect the position of the object after planning.
I am sharing a modified scene of "3DoFHolonomicPathPlanning.ttt":
https://drive.google.com/open?id=1tMgt- ... dBlTcSXSgR
The scene demonstrates the following issues:
- Changing the parents of "StartConfiguration" many times causes the position of the object to change. E.g. changing the parent 1000 times (to "no parent" and back to "Dummy") causes a positional error of 0.00046. The more iterations, the greater the error. To reproduce, enable child script of "Dummy" and disable child script of "StartConfiguration" and run the simulation.
- Running the OMPL path planner for "StartConfiguration" to find a 3D position path to "GoalConfiguration", also causes a positional error for "StartConfiguration" after planning. E.g. if planning for 10 seconds, the error is at least 0.04. To reproduce, disable child script of "Dummy" and enable child script of "StartConfiguration" and run the simulation.
What is special about this scene is that "StartConfiguration" has a parent ("Dummy") and the the rotations relative to world are not (0,0,0). I suspect that there is a calculation error for the position of "StartConfiguration" relative to its parent, either when the parent is changed, or when "StartConfiguration"'s position using global frame is changed (e.g. by the planner during collision checking).
Let me know what you think. Thank you.
Best,
Uli
I am sharing a modified scene of "3DoFHolonomicPathPlanning.ttt":
https://drive.google.com/open?id=1tMgt- ... dBlTcSXSgR
The scene demonstrates the following issues:
- Changing the parents of "StartConfiguration" many times causes the position of the object to change. E.g. changing the parent 1000 times (to "no parent" and back to "Dummy") causes a positional error of 0.00046. The more iterations, the greater the error. To reproduce, enable child script of "Dummy" and disable child script of "StartConfiguration" and run the simulation.
- Running the OMPL path planner for "StartConfiguration" to find a 3D position path to "GoalConfiguration", also causes a positional error for "StartConfiguration" after planning. E.g. if planning for 10 seconds, the error is at least 0.04. To reproduce, disable child script of "Dummy" and enable child script of "StartConfiguration" and run the simulation.
What is special about this scene is that "StartConfiguration" has a parent ("Dummy") and the the rotations relative to world are not (0,0,0). I suspect that there is a calculation error for the position of "StartConfiguration" relative to its parent, either when the parent is changed, or when "StartConfiguration"'s position using global frame is changed (e.g. by the planner during collision checking).
Let me know what you think. Thank you.
Best,
Uli