Meeting Notes: MoveIt Manipulation WG Meeting - June 25th

Attendees

  • Nathan Brooks
  • Michael Görner
  • Felix von Drigalski
  • Michael Ferguson
  • Robert Haschke
  • Mark Moll
  • Tyler Weaver
  • Henning Kayser
  • Andy Zelenak
  • Jere Liukkonen
  • Jafar Abdi
  • Marq Rasmussen
  • Dave Coleman

Agenda

  • Scene graph support (Felix)
  • Reducing compile time (Jafar)
  • When should we branch for Noetic? (Dave)
  • MoveIt2 Development update (Henning)
  • Google Summer of Code update (Mark)
  • Dropping Kinetic support (Michael)
  • Clarify 3D object import button in Rviz (Felix)

Discussion Notes

  • Scene graph support (Felix)
    • Felix: we can skip this topic, nothing to report
  • Reducing compile time (Jafar)
    • Switch to one shared library for all of moveit_core
      • Reduced compile time from 75 to 45 seconds
      • Robert: doing this isn’t reasonable
    • Should we use custom caches with build farm?
    • QT dependency large, can we pre-compile headers
    • Robert: there are less timeouts on build farm now, maybe not urgent
    • Summary: Jafar’s findings so far are not actionable due to drawbacks
    • Next: Jafar will look further into pre-compiled binaries
  • When should we branch for Noetic? (Dave)
    • Robert: Do it right before we have release process started
    • Michael / Ferguson: lets just release off of master for first 6 months while Noetic settles down
    • Robert: don’t have noetic-devel branch until December
    • Dave: do we need a noetic branch since there won’t be another ROS 1 version after Noetic?
    • Fergs: ROS 2 is not as easy as advertised. Major missing portions still (see his blog)
    • Andy: ros_control is not ported (Fergs: 6 mo still)
    • Nathan: we used MoveIt 2 for a client already
    • Summary: we won’t make noetic branches for at least 6 months
  • MoveIt2 Development update (Henning)
    • Preparing Foxy support, aiming for start of July
    • There is a compiling branch for Foxy, will have diff branch for Eloquent
    • Still working on straight migration
      • move_group almost done
      • motion planning plugin not done
      • warehouse remaining
    • moveit2 project spreadsheet shows what remains: we should be done end of July
    • need an eloquent branch because launch files changed
    • Fergs: eloquent support will be dropped soon, don’t bother with it
    • Dave: can we just drop eloquent support?
    • Tyler: there’s no foxy support in debian, but yes by source
    • Michael: we agreed not to support old ROS 2 versions
    • Fergs: rolling release is coming (Rolling)
    • Trajectory execution available in ros2 control
    • Summary: will ditch Eloquent branch and focus on Foxy
  • Dropping Kinetic support in master branch (Michael)
    • Mark: Kinetic is still widely used per the stats
    • Michael: its really hard to upgrade robots, reality is kinetic is important
    • Would need to update source install instructions to support QHull
    • Michael: we could pin the last known working version of master branch for kinetic
    • Mark: porting major features back isn’t going to happen anymore
    • Dave: having great new features in a newer release encourages people to upgrade
    • Robert: let’s drop support out of box but provide advanced qhull source instructions
    • Dave: will we turn off build farm testing for kinetic?
    • Robert: yes
    • Summary: we are dropping support for Kinetic in the master branch, only supporting melodic and noetic
  • Google Summer of Code update (Mark)
    • Jeroen student is working on OMPL constraint planning in MoveIt
    • Tyler: intern Adam is doing well working on jog_arm, addressed issues in moveit 1 first
      • Working on port for moveit 2
      • ros_control2 will be a future issue
    • Andy: renaming jog_arm to moveit_servo in ROS 1 and ROS 2
  • World MoveIt Day (Mark)
    • Still compiling statistics
    • Didn’t break any records but it was comparable to previous year attendance
  • Clarify 3D object import button in Rviz (Felix)
    • Looking for feedback on this UI change
  • MoveIt 2 Weekly Standup (Dave)
    • There is a separate more technical meeting happening if you want to get involved in hands on moveit 2 development. Contact henning@picknik.ai to get the calendar invite
  • New PR for supporting multiple planners in move_group via moveit_cpp (Henning)
    • Currently moveit_cpp supports multiple planning pipelines
    • Question: do we want to represent the planing pipeline with the planning_id? Or have separate field in the motion planning request message for the pipeline?
      • Henning prefers second option
      • Michael: do we want to allow setting a planning ID on the parameter server?
    • Feedback encouraged
    • Summary: Enforce no slashes in the planner manager ID and use separate fields for the two concepts
2 Likes

I missed a lot, thanks for putting these nicely up here.