MoveIt Maintainer Meeting Nodes - June 27th, 2019

MoveIt Meeting Notes:

Michael Gorner is back!

  • He has a 2 month old son! Congrats!

World MoveIt Day Locations:

  • Discussed MoveIt Day Locations:
    • Tokyo - Felix
    • Boulder - PickNik
    • Boston - Ian McMahon and Bryce
    • San Antonio - Jorge Nicho and SWRI
    • Germany - Michael Gorner expressed interest, but would likely just be his lab.
    • San Fransico - Dave will work on it (+ Jon Binney)
  • We’ve got a how to do World MoveIt Day (Note: don’t know where this document is, but Logistics under https://moveit.ros.org/events/world-moveit-day-2018/ is probably similar)

GSOC Project Progress

  • Jens:
    • Things are progressing, continuous collision checking via casting works. A PR is still in development, but will be open to the public soon (a few weeks)
  • Omid:
    • Has been working on another project, but will progress with TrajOpt next week.

MoveIt Example Apps

  • Led by Intel China (Yu Yun in the rep on this call):
  • MoveIt Example Apps repo exists (https://github.com/ros-planning/moveit_example_apps)
    • Handling dependencies for the CI will be difficult, timeouts are common
    • Each application can have it’s own docker image with prebuilt dependencies
      • Also open to other ideas, other contributions as well
  • Will there be examples with real hardware?
    • Each application will gurantee working in simulation
    • Make some nodes in the application tutorial
    • Will make an initial PR soon

Making a Kinetic/Melodic Release (Robert Haschke)

  • Lots of commits on master, Robert has back ported some changes back to Melodic
  • However, no one is really using Melodic, so we can’t confirm that it’s still operational and stable.
  • But we still need to release so we can stop older bugs being brought up on the issue tracker.
  • Robert will prepare a release, and has already cherry-picked bug fixes back to Melodic
    • We’ll also need a list of other commits that could go back
  • PickNik has been considering taking on more often releases from Master, likely a PPA on top of ROS. It wouldn’t need it to include the rest of ROS, and should have a different debian package name.

Benchmarking work: Henning Kayser

  • Adding planning pipeline support for the benchmark packages, simplified without warehouse, simple similarity measure for traj on the same experiment (how repeatable it is).
  • Contains all the different runs, can load different log files.
  • Could extend the similarity metric to handle different scenes as well, even though it could completely change the final results
  • A current WIP PR is out now, but reviewable PRs are coming next week.

The MoveIt 2.0 Roadmap/Future Plans

  • https://docs.google.com/spreadsheets/d/1lyLJBkPqtwdDkBAJqNf5oTO0D1cjGrO6SVM-R0yf9cY/edit?usp=sharing
  • Are the state machines in MoveIt in the right place?
    • We don’t want to write our own state machines
  • Why behavior trees vs state machines?
    • trees give you a higher level idea of what is happening
    • State machines are more event based, although trees can be as well
    • Behavior trees don’t allow simultaneous actions
  • What do we get by adding state machines?
    • allows more complicated visualizers, but also just models logic better than manually include a bunch of if statements.
  • Action hooks that send out events might become an interface, other libraries could hook into those with their own state machines.
  • Fully Leverage ROS2
    • Components nodes = 0 copying messages, etc.
    • There are a lot of work and refactoring to be done to switch: is it worth the effort?
      • the interface still must be ROS msgs, no copying be possible serializing.
      • In general, C++ interfaces are easier to use, although ROS IDL might be easier to understand
    • In terms of Grant funding, ROS 2 is desired, so it’s one of the reason’s we’re looking into it
  • Determinism and Improved Interfaces are the more aligned with what most contributors to MoveIt do.
  • We should consider what our target users are (a few examples use cases), and what our design philosphy is
    • Look for a discourse post soon

Misc

  • Sean Yen from Microsoft, have been working with ROS for a while now.
    • They’ve made a Windows version of MoveIt, more coming in July.
    • They have a CI set up already, should be as easy as retargeting the repository.
    • We have Microsoft’s support for helping resolve Window’s specific issues, there will be conversations soon about if those issues should hold up PRs
3 Likes

congratulations, @v4hn!

2 Likes

we use melodic-devel as it has all the features we currently need and is stable. In retrospect it was released at a very good point in time. Besides some random new python dependencies breaking our ci it seems to work very well. As we have a customn optimization in fcl 0.6 and updated urdfdom we don’t use the ROS packages however. I would appreciate not backporting too much stuff :slight_smile:

any update on the actual date for World Moveit Day? I have a blocker in my calendar for sometime between Nov 19 - 21? But that would collide with our ROS-Industrial training at Fraunhofer IPA, so that we could not send one of our trainers to the planned event in Hamburg…
https://www.ipa.fraunhofer.de/en/Events/ros_industrial_training.html

World MoveIt Day was announced during our maintainer meeting a while back to be November 20th! We need to more publicly announce this, for sure.

1 Like