ROS 2 Real-time Working Group Online Meeting 6 - Nov 27, 2019, 7AM PDT (UTC-8)

Hi all, below the agenda for our next meeting:

  1. Action items from earlier meeting: ROS2 Real-time Working Group Online Meeting 5 - Nov 13, 2019, 7AM PDT (UTC-7)
    1. Silexica will present on Dec 11 - Performance Testing Platform
    2. AI: Dejan to write a longer Discourse post on how to measure performance correctly
      1. Dejan did not have time
    3. AI3: @eboasson to finalize the issues for single-process, real-time rmw
  2. @ralph-lange LET executor: ROS2 Real-time Working Group Online Meeting 5 - Nov 13, 2019, 7AM PDT (UTC-7)
  3. @eboasson architecture of single-process, real-time rmw
  4. @wjwwood do you have ETA for landing of this Static Executor PR https://github.com/ros2/rclcpp/pull/873?
  5. Is anyone from from this WG active in ROS Tooling Working Group Online Meeting 3 - Friday Nov 18 3:00PM PST (UTC-8)?

Meeting coordinates:
Topic: ROS 2 Real-time Working Group Online Meeting 5 - Nov 17, 2019, 7AM PDT
Time: Nov 27, 2019 07:00 AM Pacific Time (US and Canada)

Join Zoom Meeting

Meeting ID: 599 123 651

One tap mobile
+16699006833,599123651# US (San Jose)
+19294362866,599123651# US (New York)

Dial by your location
+1 669 900 6833 US (San Jose)
+1 929 436 2866 US (New York)
Meeting ID: 599 123 651
Find your local number: https://zoom.us/u/acGuzJFesI

2 Likes

@Dejan_Pangercic

i could not make it, could you share the minute?

Best

@tomoyafujita and all, here are the minutes:

Participants

  • Christophe (Apex.AI)
  • Jaime (eProsima)
  • Ralph (Bosch)
  • Geoff (Tier IV)
  • Carlos (Apex.AI)
  • and others

Meeting recording

Meeting minutes

  1. Action items from last meeting:
    1. Mentioned that Silexica will present their Performance Testing Platform on or around Dec 11
    2. Discourse post on how to measure performance correctly
      1. 3rd point: ROS2 Real-time Working Group Online Meeting 5 - Nov 13, 2019, 7AM PDT (UTC-7)
      2. Dejan did not have time
      3. @Jaime_Martin_Losa talked about eProsima’s test setup/system
      4. AI for next meeting: follow-up on this
  2. LET executor https://github.com/ros2/rcl/pull/534
    1. rcl might not be the right place given the current architecture, so might split it up
    2. @ralph-lange mentioned that it might be a good idea to push things down to rcl so that executors could be implemented somewhere other than rclcpp while avoiding code duplication
    3. AI for next meeting: have someone from OR comment on the idea above, perhaps in the meeting directly
  3. Architecture of single-process, real-time rmw
    1. Will come back to it in the next meeting
  4. Static executor https://github.com/ros2/rclcpp/pull/873
    1. Latest update SingleThreadedExecutor creates a high CPU overhead in ROS 2
    2. PR efforts halted; will be offered as a separate package
    3. Will wait for changes to the rclcpp executor planned for Foxy
  5. Tooling WG
    1. No one from the RT WG participated
    2. Tracing tools could fit into that, @christophebedard plans to attend the next meeting
1 Like

@wjwwood @ralph-lange @JanStaschulat I see that PR in the point 2 above is a little bit at the stall point. Could you 3 maybe have a separate meeting about this and discuss where it is the best to implement LET executor? Or Alternatively, Jan/Ralph should reply to William/Jacob’s comments from 4 days ago.

@Dejan_Pangercic, @wjwwood, @JanStaschulat We will pose the PR to rclc instead of rcl. In addition, we will extend the documentation by use-cases/examples when and how to use the LET principle.