ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A answers.ros.org

MoveIt Maintainer Meeting Notes - March 26th

Here are my notes from the meeting.

Dave: We are now a working group, standardized across the ros ecosystem. Motion working group.

MoveitCon full day event after RosCon.

  • Dave: unconference event
    • instead of having a top down organizer
    • when you arrive at the event everyone proposes a topic they want to talk about for 30 min and then there is a vote which creates the topic list
    • rather than giving a talk, it is a discussion
  • Felix: lightning talk might fit better
  • Mark: we sort of had this at the workshop with postit notes. Would be good to still have some talks and poster sections too. Focus on academia and what students are doing
  • Dave: agree on focusing on what students are doing

GSOC

  • Mark: timeline has shifted due to covid-19, some good proposals, one project on cart, one on traj, several on grasping, and one on porting move_group to ros2. Now is time to mentor to help the right student to get a slot for something you care about.
    • Sort of advice needed for students is helping student plan their time
    • Formal process where you are added to the project as a mentor
    • Free intern where you
  • Felix: Where are the projects?
    • Mark: Should be visible if you go to GSOC website, if you go to MoveIt website you can find it on the roadmap. It doesn’t have to be one of those projects if the student has a better idea.
  • Felix: I’d like to do this

Enable ABI checks on bloom releases

  • Dave: is there someone where who can speak to this?
    • Amazon has made a tool for this in ROS2. Maybe we could flip a switch to enable this?
  • Henning: We are planning larger changes to moveit2, and this will cause conflicts. This should be used for stable packages only. Do we want to do develop releases while we progress in the roadmap first?
  • Dave: Does this work for ROS1.0? Just a bloom setting, we could turn it on.
  • Henning: For MoveIt this totally makes sense.
  • Dave: Will make issue on github to enable this for ros1
  • Henning: Maybe for noedic release?

melodic_devel branch not doing CI

  • Dave: issue is closed

wrong assumption about convex mesh planning issues

  • Martin: look at image, if there is a mesh if you subdivide and apply padding it is no longer the shape it was. https://github.com/ros-planning/geometric_shapes/issues/127
    • Weird behavior. v4hn commented but isn’t here.
    • Is it possible to make changes for melodic?
    • We have ideas for how to solve this problem but it could break existing code if they depend on this.
  • Dave: What is the difference between scale and padding?
  • Martin: scaling is just multiplying distances which keeps dimensions, padding is adding some fixed amount to dimensions to box. relative vs absolute number. Padding to primitive cube is different from how mesh cube.
  • Dave: are you proposing padding to always work like scaling?
  • Martin: Vertex should stay in middle of plane.
  • Mark: You can’t keep all proportions the same?
  • Martin: Can’t keep proportions, but keep it more like you’d expect. If you have a plain, you still have a plain.
  • Henning: Current implementation would allow collisions with padding. If you look at right at the edge they don’t satisfy the requirements. You could say that if you add 1cm of padding you could have collisions that are closer. Are there competing implementations?
  • Martin: Not aware. 3d modeling software only offers scaling.
  • Mark: Push out plains and scale them up.
  • Felix: Would get the expected behavior but with sharp angles.
  • Jorge: Could you compute the padding by scaling the mesh?
  • Martin: Maybe, but that would result in the current behavior or similar to proposed padding to planes.
  • Henning: Wouldn’t that be the same as removing verticy padding?
  • Martin: v4hn proposed a solution I’m not sure it would work. Behavior would change, is this acceptable for melodic?
  • Dave: Is this acceptable?
  • Martin: Can’t call this a bug, this is a specified behavior even if it is weird.
  • Felix: If the proposed solution is more intuitive and safer, then we can backport.
  • Martin: Try to figure out guarantees for backporting.
  • Marq: What is the proposed new solution? We need to pick vertices and then scale the mesh? We could average padding or take two points and scale?
  • Martin: Either you can take two vertices in a plane?
  • Marq: Are we only concerned about general shapes or only special shapes?
  • Martin: No, I think the special shapes only show this problem?
  • Mark: Calculate based on how sharp the normals are. Sounds complicated.
  • Marq: If I asked for scaling looking at the object I’d want it along the fate normal. I would want to add padding along the 0 line, not along the 45 angle.
  • Martin: I think that is what we proposed.
  • Henning: That means taking the normal of each vertices and have new vertices that aline with new normals.
  • Martin: Take all padding planes and get the new vertices
  • Henning: Isn’t it costly to compute?
  • Dave: Look forward to pull request Martin

Dave: our ability to write unit tests could improved with more tools

  • Felix: are you proposing test
  • Michael: There is one package where you can import the Panda or an ascii one
  • Dave: Brice contributed to tutorial a robot builder tool. How can we improve test coverage.
  • Michael: This seems like looking for a framework to solve a people problem.
  • Dave: Worked on project with good tools.
  • Michael: The right way to approach this is to write more tests and from that we’d discover missing tests.
  • Tyler: We should just write more tests

Any new updates on Moveit 2.0

  • Henning: Development has paused, not any real feeding
  • Michael: Is is paused because funding
  • Dave: Yes, funding, didn’t get ROSin because of EU regulations

World MoveIt Day (choose date)

  • Dave: Mark was looking at conference cal, Oct 1 or Oct 8.
  • Mark: Some time between WMD and MoveItCon so we can announce progress, not just a few days.
  • Felix: Won’t be able to host anything before middle of Oct, 14-15 is the earliest
  • Mark: Not near paper deadlines, if we do it on the 15th.
  • Felix: Will be at competition with robots (world robot summit)
  • Dave: Is it robot arm
  • Felix: It is well funded by Japanese government, industrial robots, 60-70 teams.
  • Mark: That sounds like a good reason to move it.
  • Dave: Earlier?
  • Mark: That conflicts with deadlines for students
  • Michael: What about after MoveitCon?
  • Mark: Would like to announce results.
  • Michael: Maybe after? It might make sense to talk about issues there and then announce the day there.
  • Dave: What about in Jan?
  • Mark: Conflict with students.

Michael: Thank you Martin for all your contributions

2 Likes

@tylerjw Thank you for putting this nicely here.

Here’s a link to the World Robot Summit, which runs from Oct 8-11 (for most of the challenges) and Oct 20-22 (for the Disaster area in Fukushima).

I would suggest early or late November for WMD myself.

1 Like