Meeting Notes - 22 April 2021
GSoC – student applications, proposals, mentors
Release updates – Noetic, Melodic, Galactic, backports
Diff Drive motion planning update
Tyler pitch related to Melodic discussion
Lots and lots of proposals!
Comment on the PDFs themselves within the GDrive folder (a link to this folder was emailed separately to meeting participants).
Strong applicants have strong/relevant backgrounds, reasonable plans, prior experience with ROS and manipulation
MoveIt releases (Mark)
Discussed in order of oldest to newest.
Melodic – Michael (not present, skip discussion for now)
Noetic – Tyler
Until catkin_package is released for Noetic, a lot of stuff is blocked for Noetic.
Let Tyler know if things have been merged in that isn’t ready to be released
Dave: Want to make sure we actually get releases out.
Tyler: vcs_tool is distributed through package_cloud (alternately, in addition to debians and pip)
Free for open-source projects, provides the backing infrastructure for supporting around-the-world reliable debian repos.
Use Github Actions to create a deploy step that deploys debians to PackageCloud and provide instructions for users to add a PPA to use these releases explicitly.
Lets us release more regularly without requiring that MoveIt stays API-stable – reduces friction.
Another question we need to answer: can we push out multiple release versions?
Mark: Possibility for having conflicting/incompatible debians for various dependencies.
Tyler: Another option is conda, though would prefer to use a PPA instead (nicer experience for users)
Robert: Debian packages can in principle handle different kinds of interdependencies, but it partly depends on what Bloom is doing to set this up currently.
Tyler: In the long run, we’d use this approach to allow people to use updated API for older distributions without us needing to push breaking changes to the official releases. Wondering if work on providing backports if really worth it (does that user really exist, or are there actually enough of them?)
Dave: Tend to assume that those users don’t exist if we’re not hearing from them.
Felix: Likes the idea of separate API-stable and cutting-edge releases.
Tyler: Alternative (but self-admittedly dubious) idea: automatically create separate branches for “future” releases, make it clear in the README that they’re not stable, and release through the buildfarm. Mirrors what rosbag group had to do to get around breaking API issues for Foxy.
Dave/Mark: Probably a good idea to schedule a separate meeting for this topic. Return to at some point in the future.
Foxy release – failed in buildfarm, putting out a patch to fix it.
Galactic release in May – what’s left to do?
Switching to compiling the main branch on Galactic. Most dependencies have been ported over.
Dave: How are we actually going to execute the release? Website says we have a feature freeze on May 30th.
Henning: Turn on CI for Galactic on the freeze date (and require it to pass), then focus on fixing bugs that come up until the release. Foxy to Galactic is a smaller step than Eloquent to Foxy, so probably won’t need many #ifdef types of things.
Tyler: will spearhead turning on CI for Galactic.
UR drivers are in beta and are ready to use. Going to transfer from the PickNik organization to the Universal Robots organization
Pilz Planner Port (Henning): WIP, but it’ll take time to get all the tests enabled. Would be great to get it added to MoveIt2.
MoveIt Setup Assistant port: major effort to get it to be useful.
Dave: Windows support?
Tyler: Need to revisit this. Depends on changes in PRs that are now stale.
Lou: In the process of updating PR and bringing in some new changes for Windows support. Also looking into Github Actions support for Windows build. In the longer run, looking at Windows realtime support with ROS2.
Dave: Making CI builds non-optional should be the goal (otherwise we’ll just break it). Good to know about things to watch out for to make sure we can support Windows.
Tyler: Can also make a special branch for Windows where changes need to be ported in.
Hello Robot support
Mobile manipulator with a diff drive base, and prismatic joints for the arm and torso.
Needs base to move around during manipulation to get full 6DoF motion
Created a new diff drive joint that lets us do integrated planning for the arm and base together using MoveIt. Coming in MoveIt2.
Showed some cool video of the robot driving through some narrow gaps to plan a grasp.
This is not a replacement for Navigation stack, but it’s useful for planning coordinated manipulation.
Will email out a link to a document to let people comment on GSoC proposals.
Dave: At the next meeting, would like to devote significant time to updating the MoveIt roadmap.
Dave: Feedback from non-PickNik participants on the call? Should we do something different to get more engagement from the open-source community?
Robert: Announce meetings earlier, or use additional channels.
Mark: Was a little late this time with the Discourse announcement (8 days before).
Felix: Discord is actually quite active, so we should announce there as well. Ties into engagement with new members – topic for the future!
Abishalini Sivaraman, Andy Zelenak, Cassidy Elliott, Dave Coleman, David Lu!!, Felix von Drigalski, Henning Kayser, Joe Schornak, Jafar Abdi, Lior Lustgarten, Lou Amadio, Mark Moll, Marq Rasmussen, Nathan Brooks, Robert Haschke, Simon Schmeisser, Tyler Weaver, Vatan Aksoy Tezer