Technical Steering Committee (TSC) Meeting #9 Minutes

Autoware
1

I apologise for the lateness of these minutes. The meeting took place while I was on holiday and I forgot to add posting them to my todo list for when I got back.

August 15, 2019, 22:00 UTC

Attendees

  • Arthur (Huawei)
  • Lee Baldwin - AutonomouStuff
  • Geoffrey Biggs (Tier IV)
  • Esteve Fernandez (Apex.AI)
  • Kenji Funaoka (Tier IV)
  • Shinpei Kato (Tier IV)
  • Seonman Kim (LGE)
  • Kiichiro Okada (Macnica)
  • Tsutomu Otake (Macnica)
  • Paul Sastrasinh (TRI-AD)
  • Antonis Skardasis (StreetDrone)
  • Stephane Strahm (Kalray)
  • Tomoyuki Watanabe (Macnica)
  • Josh Whitley (AutonomousStuff)
  • Dmitry Zelenkovskiy (LGE)
  • Yang (96Boards)

Minutes Geoffrey Biggs (Tier IV)

Agenda

  1. Opening remarks and new member introductions (Board)
  2. Admin Confirmation of previous minutes (All)
  3. Follow-up Action items from previous meeting (All)
  4. Admin Hiring a software architect
  5. Admin Names and photos of TSC members for the AWF website
  6. Admin Autoware annual summit update
  7. Admin Access problems from China
  8. Working group report Autoware (@gbiggs, @esteve)
  9. Working group report Map formats (@Brian_Holt)
  10. Working group report Vehicle interfaces (@Antonis_Skardasis, @JWhitleyWork )
  11. Working group report ECU/Platform (@sstrahm)
  12. Working group report Simulation (@bshin-lge)
  13. Other topics: Autoware racing (@Antonis_Skardasis)

Action items

Detailed minutes

Opening remarks and new member introductions (Board)

  • Members are joining on a weekly basis. We still have challenges to solve, in particular in getting contributions, but it shows that we are doing well.

Confirmation of previous minutes (All)

  • Minutes approved.

Follow-up: Action items from previous meeting (All)

  • Release the safety case example
    • @Brian_Holt
    • Action taken Still waiting on the people on the other end to move it through their legal team. It is possible that they will choose not to release it.
  • Start a Discourse poll on when to hold the Autoware Summit
  • Begin scoping the work and skills for an autonomous driving system architect

Hiring a software architect

  • The Autoware Foundation is hiring a system architect to lead the design of Autoware.
  • Apex.AI has posted a job notice already.
  • Tier IV is looking into posting it on its careers page.
  • We would also like Linaro to post it.
  • The Autoware Foundation should also consider how to further spread the notice.
    • Post it to the ROS Discourse.
    • Post it to robotics worldwide mailing list.

Discussion

  • Linaro is willing to host if necessary. Add Cambridge to the list of possible locations.

Names and photos of TSC members for the AWF website

  • The list of TSC members on the foundation website is out of date.
  • If you are not listed, please send your photo and name so we can get you listed.
  • Only the primary TSC representative should be listed

Autoware annual summit update

  • The annual summit was proposed to be held in Macau just before or after ROSCon (end of October or beginning of November).
  • Due to a low number of potential attendees, this proposal has been dropped.
  • A teleconference summit will be held instead.
  • Instead of the physical summit, the board encourages local meet-ups.
    • Parkopedia notes that they are holding a three-day hackathon in London at the end of September. Linaro will also be participating in this.

Access problems from China

  • We have recently received reports that Autoware users in China are having trouble accessing the Autoware repositories hosted at GitLab.
  • Investigations have found that this is a sporadic problem outside the control of the Autoware Foundation or Gitlab.
  • Possible solutions considered so far:
    • Shifting back to GitHub
      • We have found that GitHub is also sporadically blocked in China, so this will not solve anything.
    • Hosting mirrors at GitHub
      • When we created mirrors at GitHub for the GitLab repositories, we immediately started getting issues posted to them, so it is clear that having two sources of the code is confusing for users.
      • This also suffers from the “GitHub is sporadically blocked” problem.
    • Creating local mirrors in China
      • Not yet investigated, but Linaro has used this solution in the past.
      • There will be some cost for this; hopefully the foundation can cover it.
  • In the short term, a post has been made on Discourse indicating we are aware of the problem and investigating solutions to it.

Working groups process update

  • A wiki for the working groups has been created.
  • The leaders of all working groups must fill in a wiki page for their working group.
  • The wiki page should provide:
    • Who the leaders are and how to contact them (Discourse PM, via GitLab, etc.)
    • An overview of the working group’s goal(s).
    • The current status of the working group.
    • A list of links to the working group’s meeting minutes.
    • Any other relevant information.
  • Working group leaders are required to keep their WG’s page up to date, especially with the list of minutes.
  • Look at the page for the Autoware working group for an example format.

Working group report: Autoware (@gbiggs, @esteve)

  • The dates for the next release of Autoware.AI have been set by the maintainer team.
  • Autoware.Auto updates:
    • The first Autoware.Auto first release has been submitted to the ROS build farm so that binaries will be available.
      • Some issues were found during the binary building so it has been delayed for one sync cycle.
    • The ros1_bridge work is moving forward, with the bridge being added as a dependency of Autoware.Auto to avoid making ROS 2 a dependency of Autoware.AI.
    • The new velodyne driver developed for Autoware.Auto is being pushed up to the ros_drivers repository. This work is also being used as a basis for a new framework for all ROS drivers that encourages a more standardised driver structure.
    • Epics for the first milestone have been created and broken down into tasks. The Autoware working group is working on assigning these tasks.

Working group report: Map formats (@Brian_Holt)

  • Have been meeting every week.
  • Discussed physical storage formats.
    • Consensus was that OpenDrive and OSM XML are the only realistic formats.
    • In the short term, will use OSM XML with the Lanelet2 library.
    • In the long term, will use OpenDrive for the physical storage format.
    • Lanelet2 will be the API used to access map data, because it is the best option available. However it is not perfect; it needs modifications and extensions. We will try to make these modifications upstream to avoid needing to create a fork.
  • Discussed simulators, specifically with regards to supported map formats and environments.
    • There was less consensus, with all considered simulators having pros and cons.
  • Next meeting will discuss map use cases.
  • Meetings will soon shift to a biweekly schedule to give more time for work between meetings.
  • The WG is aiming to have support for Lanelet2 and OSM XML in the next release of Autoware.AI, although the Aisan vector map format will still be the default. For the following release, OSM XML/Lanelet2 will be the default, and in the release after that they will deprecate the Aisan format.

Discussion

  • AutonomousStuff: The maintainer team had some problems getting Lanelet2 to build, but these have been resolved. Lanelet2 is now included in the default Autoware.AI distribution.
  • AutonomousStuff: How are these large changes going to be made? Who will do them?
    • For Autoware.AI, it is a plain old OSS project. If you want something done and do it, then it will get done. If you don’t, it might not.
  • Tier IV: Will Lanelet2 fit all the use cases as an API, or are there likely to be use cases where it won’t work well?
    • For the most part it should be representative. We are hopeful.

Working group report: Vehicle interfaces (@Antonis_Skardasis, @JWhitleyWork)

  • The last meeting discussed the physical layer protocols.
    • Determined that most members are using only CANBus so that will be what is dealt with.
    • There is a set of messages for CANBus from ROS Industrial, and members generally felt positively about these.
  • The next step is to identify the CANBus usage of each vehicle to understand what each vehicle provides, which will hopefully lead to a holistic view of CANBus messages covering all target vehicles.
  • They will move on to ROS messages after that.

Working group report: Reference Platform (@sstrahm, @cheng.chen)

  • First group meeting was held earlier today, but was only attended by the core members.
  • There is a very large amount of interest in the creation of a reference platform, as a hardware platform for running Autoware.
  • The target of the working group was discussed.
    • The WG will not define a single reference platform.
  • The first reference platform will be from Linaro/96boards and Autocore.
  • The WG will look at defining the prequisites and requirements for a platform, such as performance and how it will be tested.
  • The WG notes that it must remain neutral on the selection of hardware, even though the work may be done by people from companies who provide boards. Hardware from outside the foundation must also be given fair consideration.
  • The WG notes that it will need to have a strong relationship with Autoware.Auto and the Autoware working group.
  • The WG wants to have a commitment to having a good out-of-the-box experience for users so that they have a good first-time experience with Autoware and their chosen reference platform.

Discussion

  • StreetDrone wants to be considered as a supported platform.
  • Parkopedia: Is there a plan to support the NVIDIA PX2?
    • That is something the working group will discuss.
    • It is difficult to do with NVIDIA not being a member, but on the other hand it is a commonly-used piece of hardware.
    • We are not excluding any platform as an open source project, and the PX2 is likely to continue to be used by members, but we need to promote member platforms foremost.
    • AutonomousStuff: NVIDIA is no longer offering or supporting the PX2 because the Xavier platform is now available.
  • Apex.AI: Through the ROS 2 real-time working group, we are working on running their build performance test on the buld farm.

Working group report: Simulation (@bshin-lge)

  • The working group has just made its call for participation and has not yet held its first meeting.

Other topics: Autoware racing (@Antonis_Skardasis)

  • There is increasing interest in autonomous vehicle racing.
  • The founder of StreetDrone (Mark) is involved in car racing and has proposed an autonomous race. He has attracted interest from OEMs.
  • Autoware.race?
  • There is already a plan for seven races in three different countries, to occur sometime in 2020.
  • The point is to test algorithms in a safer environment than public streets.
  • More details cannot yet be given because they still require an NDA.

Discussion

  • Will the car or the software vary?
    • Currently it is planned to have a single type of car (StreetDrone’s platform) and for the software to vary by team.
3

@gbiggs thx, thorough as always.

Couple of remarks:

Above page does not exist.

Also this does not exist.

Will @Antonis_Skardasis still be leading VI group?

Is this group aware of this proposal for vehicle interface abstraction Add vehicle interface for Autoware · Issue #316 · autowarefoundation/autoware_ai · GitHub ?

@sstrahm What is the definition of platform here? Just ECU? Is vehicle included?

Apex.AI is for our internal purposes investing into a solution that would allow us to flash, provision and run automated (performance) tests. The solution is being built around The Custom executor | GitLab and we will support:

  1. Renesas R-Car H3
  2. Nvidia Jetson AGX
  3. Nvidia Drive AGX
  4. Yocto Embedded Linux
  5. QNX

We plan to also add a support for test result storage and visualization. If this group is interested into a system like this please contact me. There is currently big interest from ROS 2 community for such system and even commercial companies want to get involved.

We are also still looking for someone that could host such a farm of such platforms and continuously run the the tests. We are currently talking to https://baylibre.com/ and https://www.osadl.org/.

1 Like
4

Sorry, they were moved after these minutes were created. I have fixed the links.

5

Will @Antonis_Skardasis still be leading VI group?

For the time being, I will be leading the VI group. The TSC will be discussing getting me a co-chair. Our current focus is the Vehicle Interface Community Survey - #2 by JWhitleyWork.

And no, I was not aware of that proposal document. Thanks for the heads-up! We will review it in the next meeting.