Autoware Working Group meeting 20191002

Autoware
1

The next Autoware working group meeting will be held on Wednesday, October 2, 2019 3:00 PM.

Please join if you have an interest in contributing to the next generation of Autoware.

Here is the current agenda:

  1. Report on the Parkopedia hackathon
  2. Status of work on Autoware.AI 1.13 release
  3. Go over the list of MRs to be reviewed and merged into master for 1.13.
  4. Status of the Autoware Specification work and discussion of the proposed specification creation process

Please post any additional discussion items if you have them.

Please join my meeting from your computer, tablet or smartphone.
https://global.gotomeeting.com/join/547575093

You can also dial in using your phone.
United States: +1 (646) 749-3129

Access Code: 547-575-093

More phone numbers
Australia: +61 2 9087 3604
Austria: +43 7 2081 5427
Belgium: +32 27 00 6378
Canada: +1 (647) 497-9391
Denmark: +45 32 72 03 82
Finland: +358 923 17 0568
France: +33 170 950 594
Germany: +49 692 5736 7317
Ireland: +353 15 360 728
Italy: +39 0 247 92 13 01
Netherlands: +31 207 941 377
New Zealand: +64 9 280 6302
Norway: +47 21 93 37 51
Spain: +34 932 75 2004
Sweden: +46 853 527 836
Switzerland: +41 225 4599 78
United Kingdom: +44 330 221 0088

New to GoToMeeting? Get the app now and be ready when your first meeting starts:
https://global.gotomeeting.com/install/547575093

2

Attendees

  • Geoffrey Biggs
  • Akihiko Tsukuda
  • Akihito Ohsato
  • Brian Holt
  • Esteve Fernandez
  • Filipe Rinaldi
  • Josh Whitley
  • Liyou Zhou
  • Ruffin
  • Ryohsuke Mitsudome

Minutes Geoffrey Biggs

Discussion

Report on Autoware Foundation Hackathon hosted by Parkopedia

  • Around a dozen attendees for the first couple of days, fewer on the third day.
  • Main achievement was getting a bunch of developers started on using ROS 2, the ADE, etc.
  • Parkopedia produced a package for PID control
    • Useful for longitudinal motion and steering control
  • Christopher from Apex.AI has started planning work on the pure pursuit controller
  • Once we have this, we just need NDT (which is in progress) and the waypoint follower.
  • The minimal stack that Parkopedia has produced was ported to Melodic, which means they can start using the bridge.

Autoware 1.13 status

  • Merge request cut-off was just over 24 hours ago.
  • 100 merge requests done so far.
  • 60 additional active merge requests currently
    • Two thirds have been assigned
    • One third have been reviewed
  • Major feature is lanelets2
  • There are some concerns about the balance of activity between Autoware.AI and Autoware.Auto
    • The AWF is aware of this problem and actively working to address it

Process for getting the remaining MRs reviewed and merged

  • All MRs have at least one committer assigned
    • Committer’s job is to be the final gatekeeper, judging quality of reviews done by contributors and the response of the MR submitter, and do the final merge
  • Technical review will need to be done by someone who has an understanding of the technical aspects (algorithms, etc. as relevant).
    • Most important aspect being “does it actually work”
  • Quality review can be done by:
    • The technical review person
    • A committer
    • Any one else who cares enough to do it
  • Quality review needs to look at aspects such as is it tested, is the coding style compliant with the style guide, etc.
  • Merge review guidelines (for all)
  • MR acceptance guidelines (for committers)
  • Deadline for getting MRs merged into master is 1st November so we have enough time for real-world testing before the release

Other items

  • @JWhitleyWork has just about finished porting ros-canopen (from ROS Industrial) to ROS 2, so it should be available soon for use in Autoware.Auto.
  • Data logging feature
    • AutonomousStuff has a hardware blackbox produced by Intrepid Systems that records the car’s CAN network, the user’s (drive-by-wire) CAN network, and two video signals from inside the car.
    • Kvaser has an interested in developing a blackbox data logger with CAN interfaces, because loggers is one of their specialties.
    • @ruffsl is interested in contributing to the data logging feature as a way to spread the workload for that concept in general ROS usage around.