ROS 2 TSC Meeting Minutes: 2021-03-18

General
1

ROS 2 TSC Meeting Agenda - 2021-03-18

ROS 2 TSC Contribution Report - 2021-03-18.pdf (123.6 KB)

  • Attendees
    • Open Robotics - Chris Lalancette, Kat Scott, William Woodall, Brian Gerkey, Geoff Biggs, Louise Poubel, Tully Foote
    • Toyota Research Institute - Ian McMahon
    • Samsung Research America - Steve Macenski
    • eProsima - Jaime Martin Losa
    • ADLINK - Joe Speed
    • Sony - Tomoya Fujita
    • Canonical - Sid Faber
    • Amazon - Aaron Blasdel & Emerson Knapp
    • Apex.AI - Lyle Johnson
    • ROBOTIS - Will Son
    • Microsoft - Sean Yen, Lou Amadio
    • SwRI, representing GVSC - Jerry Towler
    • Wind River - Rob Woolley
    • Bosch - Karsten Knese
    • PickNik - Dave Coleman
    • Intel - Harold Yang
  • Preliminaries
  • Old business
    • [Gerkey][2 mins] Updating membership tenure policy
      • Scheduling meeting
  • New business
    • [Blasdel][2 mins] rqt is running out of active maintainers
      • Do we need a call for maintainers?
        • Core is maintained but many plugins are not
      • Post to Discourse with enough info for volunteers to understand what’s required
        • Potentially also organize a meeting among a subset of TSC and Tooling WG to discuss options for making rqt more easily maintainable.
    • [Karsten][5 mins] status of documentation
      • How can projects such as ros-control link their documentation to docs.ros.org?
        • Can find a way to link into general docs
        • A dedicated package for tutorials is an option that will go through per package documentation (in process)
    • [Towler][2 mins] Curious about progress/status of the Rust WG (Topics tagged wg-rust)
      • Don’t think it’s a formal ROS 2 WG, so it doesn’t get a recurring slot
        • It would be good to have an update. Geoff, maybe invite Ruffin to present at TSC in updates section.
        • Blocker of cmake based message generation is in the process of being resolved for bazel integration.
  • Recurring business
    • Next ROS 2 distro release
      • Galactic
        • API and feature freeze is April 5 (2 weeks and 1 day away)
        • Big ticket items still outstanding:
          • EventsExecutor (iRobot & Open Robotics)
          • Switch to new Connext RMW (RTI & Open Robotics)
          • rosbag2 /clock topic (AWS)
          • Per-package documentation (Open Robotics)
          • rclpy → pybind11 (TRI & Open Robotics)
        • Roadmap: Roadmap — ROS 2 Documentation: Rolling documentation
    • Working groups [5 mins each]
      • [Knese] Control
      • [Speed] Edge AI
      • [Martin Losa] Embedded
      • [Coleman] Manipulation
        • Lots of stuff merged in on WMD
          • Very creative lightning talks, interest talks at the fireside events
          • Discord is a popular spot
          • 81 pull requests merged!
        • CI continues to be an issue
        • Lots of new information for beginners added to MoveIt website
        • Gave up on “first timer issue” list as MoveIt is too complicated, switching to simply “Good First Issue” labels instead
        • Still struggling to convince MoveIt community to switch to ROS 2, community segmentation happening
        • MoveIt2 development continues, several major applications have been built on ROS 2 now
        • Creating future roadmap for MoveIt
        • MoveIt in GSoC program again this summer
      • [Macenski] Navigation: Theme: “Galactic Push”
        • SLAM Toolbox 360 lidar patch!
        • 6 new contributors in the last month! Several with multiple new PRs and projects
        • Working on resolving pre-Galactic release API changes in rclcpp and merging in any ABI breaking changes we want in from Nav2 before Galactic
        • Doxygen push to get all header files covered by end of March!
          • Only 2 packages left! :slight_smile:
        • Developments on the state lattice planner and speed up collision checking on costmap 2d
        • ROS2 OpenVSLAM port and interface quality
        • Work on development of new “NavigateThroughPoses” action
          • NavigateToPose → Go to a pose
          • NavigateThroughPoses → Go to a set of poses, without stopping within some reasonable path following tolerance “viapoints”
          • Waypoint Follower → Go to a set of poses, stop, execute some task, and continue to next
        • TEB Local Planner: Christophe has given Steve admin access to TEB to maintain and working with Guillaume Doisy @ Wyca Robotics to get it back to a good state
        • V1.0.0 · GitHub 2 tickets from Nav2’s V1 we hope to have alongside Galactic.
      • [Andrei] Real-time (â„… Rob Woolley)
        • Aggregated real-time ROS2 related information: tests, docs, utils and put links to it on the main working group page
        • Created a real-time ROS2 WG chat on matrix
        • Setup two HW targets: one Intel x86_64, another ARM (everything is manual now)
          • Build RT_PREEMPT LTS 5.4 kernel for both targets, fixed some problems, since Ubuntu has moved to the 5.8 kernel in 20.04.2
          • Installed Ubuntu 20.04.2
          • Installed ROS2 rolling and foxy releases
          • Run several real-time tests.
        • Open questions:
        • Next steps:
          • Add a description of how to configure ROS2 RT system (it might be two different descriptions: demo and tests)
          • Make a kernel deb packages release, that users can grab them from the real-time community subproject
          • Create a Docker container for the x86_64 build
          • Select real-time tests to execute
        • Proposal
      • [Biggs] Safety
        • Re-working the Safety WG to focus more on information dispersal rather than building stuff. As such, we will be shifting to a once-a-month schedule and try to have a talk from someone using ROS in a safety-critical way each month.
        • The first talk will (probably) be from Shawn Schaerer and will happen in April.
        • We will also be setting up a Matrix channel for communication in between meetings, at the request of WG participants.
      • [Faber] Security
        • Continued work on implementing security for MoveIt2 and RMF
        • Not a lot of momentum; changing frequency of WG meetings to monthly.
      • [Knapp] Tooling
        • WG notes doc
        • RMW_DURATION_UNSPECIFIED / RMW_DURATION_INFINITE merged in RMW and top-tier RMW implementations - creates a consistent interface promise - before there was inconsistent behavior
          • This was needed for rosbag2 to successfully handle QoS profiles across implementations
        • Rosbag2 Foxy backport - final decision made to create a “rosbag2-future” package in Foxy. Will have similar feature set to Galactic release of rosbag2. See https://discourse.ros.org/t/fast-forward-merging-rosbag2-master-api-to-foxy/18927
        • Rosbag2 removed dependency on rmw_fastrtps - implemented more generic converter that searches installed RMW implementations for serialization format support
        • Design review wrapping up on rosbag2 handling of clock/simtime - new design allows for very flexible time control on bag playback. Implementation to begin immediately.
        • Many usability improvements in Github CI Actions - thanks to Christophe Bedard. Most notably:
          • action-ros-ci: Ability to pass arbitrary colcon configuration in code for full build&test control
          • action-ros-ci: Ability to provide linked PRs and extra repos files in PR description
      • [Woodall] Middleware
        • Meeting Notes:
        • Continuing work and discussion on outstanding design changes:
          • Event Executor and related API changes
          • Content filtered topics
          • Unique Network Flow IDs
        • Unique Network Flow IDs
          • After discussing, we’ve decided to move forward with this approach, even though there is a dissenting opinion that this information (either the request for the unique network flow in the ROS QoS, or the exposed network endpoint information) was not appropriate at the ROS level, and instead it should be handled internally to the middleware. We decided that it wasn’t an “either/or” situation and that implementing the current proposal moves us forward without precluding the other approach.
          • Pull requests still need reviews.
        • rmw_time_t replacement was discussed
4 Likes