0
$\begingroup$

I am attempting pose target planning and for any but trivial tasks it fails to find a path even with large time limits. Does it require a supercomputer? What is the expected minimum HW to solve it?

My setup:

  • ROS2 Humble, Moveit2.
  • 6dof arm (uArm Lite6) simulation
  • Ubuntu 22.04
  • VMware Workstation 17 Pro, allocated 9GB memory and 4 cores out of my 2.5GHz i5-14400F
  • RRT, RRTConnect, EST planners, all fail with 5 minute limit. RRTConnect failed with 30 minutes. All by timeout, saying they couldn't solve it.
  • It fails for a random valid target position 10cm away
  • It finds a plan when the target is the starting position, or it differs from it by 1 joint movement. (in other words, the setup does not have a trivial error - it works.)
  • Workspace bounds are reasonable; other params default. Tried changing them, did not help.
  • It works well and fast for joint target planning.

EDIT: I am not asking whether my specific parameters are correct - I am aware I have not provided them. I am asking whether a 6 dof arm pose planning requires a super computer or my hw should be able to handle it. (Consequence for me is that if yes, my parameters are likely the problem)

Improve this question
$\endgroup$
3
  • $\begingroup$ Things like motion planning the details of things like parameters are important. And saying things are "reasonable" is very subjective depending on the person and situation. Without the question being more specific the only answers you can get will be conjecture and guesses which will rely on luck to solve your problem. Please take a look at How to Ask & tour for more information on how stack exchange works and work through the Robotics question checklist to edit your question to make it clearer. $\endgroup$ Commented Sep 19, 2024 at 3:34
  • $\begingroup$ That is incorrect. This information is sufficient in order to give me answers that I want. I am not asking "How long will it take given this set of HW and parameters". I am asking "Is there any world in which this works?" I expect an answer such as "On this HW it should take 0.5 secs, your params are wrong" or "Do not expect it to work fast". Now, please reopen the question cos it took me a long time to put together. $\endgroup$ Commented Sep 21, 2024 at 9:08
  • $\begingroup$ Please remember that people here only know about what you write. We can't comment on your parameters because you haven't shared them. This can includes your success criteria, search precision etc. You may have a very hard problem or you may be approaching it in a sub optimal manner. I'm general performance on any given task is highly individual to the use case. Please read the links I sent about how to ask a good question on this site. In particular be specific and make it relevant for others sections. And consider a minimum reproducible example. $\endgroup$ Commented Sep 21, 2024 at 18:39

1 Answer 1

Reset to default
0
$\begingroup$

The simple answer to your question: No a super computer is not expected to be needed for 6 dof planning. The original RRTConnect paper from 2000 planned in 12.5 seconds on a computer from that era: https://www.cs.cmu.edu/afs/cs/academic/class/15494-s12/readings/kuffner_icra2000.pdf

Rewording this to be helpful for others. How can I find out if my algorithm is solving my problem in an expected amount of time?

To answer this question you want to find publications of benchmarks for similar problems and how others have solved them, and shared their performance. As this isn't my project I just found a quick reference for comparision by searching and pointed to that. But if you want to do this robustly you should find multiple examples and compare them. Most projects have performance benchmarks that they compare over time which can be good references too.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thank you, appreciated. I was trying to find some quotes but I failed, which is why asked here. This info puts me back on track with the work. $\endgroup$ Commented Sep 25, 2024 at 19:31

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.