Technical Steering Committee (TSC) Meeting #35 (2021/10/20) Minutes

Autoware
1

2021/10/20 15:00 UTC

Attendees

  • Ryohsuke Mitsudome (TSC Chair)
  • Shinpei Kato (AWF)
  • Yang Zhang (AWF)
  • Christian John (SPC)
  • Josh Whitly (AWF)
  • Hiroshi Igata (ODD WG)
  • Kerem Par (Adastec)
  • Filipe Rinaldi (Arm)
  • Stephen Li (AutoCore)
  • Michael Li (ITRI)
  • M. Fatih CIRIT (Leo Drive)
  • Seonman Kim (LGE)
  • Naoyori TANZAWA (Tier IV)
  • Adam Dabrowski (Robotec.ai)

Minutes:

Action items

Board Passdowns (5min)

  • 2 FTEs offered for software architect
  • 1 FTE for software developer
  • SPC topics:
    • ROS world happening this week
    • IAC: AWF preparing presence at Tech Showcase
    • Website update ongoing to update Autoware.Auto project page and Autoware.IO Product Promotion Portal
    • ARM Dev Summit: Shinpei and Girish giving a talk about Open AD Kit tomorrow

Follow up with the new Software Architect (Operation Committee) (5min)

  • Official call for the Software Architect was shared via email.
  • Fatih (LeoDrive) and Xinyu(Tier IV)
  • Starting from November 1

Follow up with the next ODD Voting(TSC Chair) (5min)

  • Bus Public: 3
  • Bus Limited:2
  • Campus Private: 2
  • Abstain: 12

Working group report: Autonomy Software (10 min)

  • Feature development for Cargo ODD: Tracking modules are almost finished. The merge requests for planners are going to be ready soon. Many merge requests are blocked by flaky CI tests. We have some members who are working to fix them
  • Preparation for Cargo demo with Eve Autonomy: it is progressing with a small delay, but we are still targeting to shoot demo video by the end of November. The vehicle hardware became ready today, and Tier IV will be going to the Ryuyo warehouse next week for the first integration. Ryohsuke will be preparing an integration branch with open MRs merged.
  • Preparation for Zing demo: They have difficulty with finding an engineer to test vehicle interface. They would like to have someone who can help them integrate Autoware on the site. If there are any volunteers, please contact Ryohsuke or Bonolo.
  • We have also started to organize launch files for the Open AD Kit. Further discussion between the WG to get aligned with tasks that needs to be done before the release of Open AD Kit 1.0.

Discussion:

  • Fatih: How many organizations are using Autoware.Auto at the moment actively?
    • It is hard to make decisions on architecture change, legacy support, etc unless we donā€™t know who is using Autoware in production.
    • Autoware is OSS so there could be any companies that are customizing Autoware.Auto for their use. If there are any significant change in the upstream repository, itā€™s distributorsā€™ responsibility to pull those changes.
    • E.g. If there are no users who strictly need Foxy, we donā€™t want to waste resources to maintain Foxy in Autoware.Auto after switching to Galactic.
    • Action Item: Fatih + Xinyu + Mitsudome to set a call to discuss this in more detail and bring it back in the future TSC call.

Working group report: Operational Design Domain (10 min)

  • Brushed up the ODD definition table of the bus service based on ITRIā€™s public road autonomous bus service project in Hsinchu-city, Taiwan.
    • Over all route length is 5.3km (1st year) and 8.1km (2nd year)
  • Discussed the details of the Seattle Tacoma airport ODD.
    • Zing Robotics joined and provided update of the indoor/outdoor maps
    • Potential use of the ceiling information for localization (indoor)
    • Highly reflective glass/metal walls can be an issue.
    • The maximum speed is 25 mph outdoors, 5 mph indoors
    • Swift handover from the point cloud localization to the RTK-GPS is required
    • Identified additional difficult scenes like the bi-directional lane
    • The vicinity of aircrafts is excluded from the demonstration scope
    • The 3D outdoor digital map was created of SeaTac airport (from the sparse and noisy point cloud data) by Tier IV and Robotec.aiā€™s contribution
  • Discussed the ODD associated issues with racing projects (F1TENTH, IAC)
    • A JSAE event ā€œJapan Automotive AI Challengeā€ was introduced as a reference use of simulation. In this event, Autoware.Auto, SVL simulator and IMS course data are used. Tier IV is supporting it with help from LG.
  • The CI/CD pipeline will be introduced to ODD working group as well as other working groups
    • The Scenario editor and the scenario runner are integrated with this pipeline
    • Will be able to test run (preview) the scenarios with Autoware.Auto running

Discussion:

  • Shinpei: More than 100 participants registered to Japan Automotive AI Challenge. We will share the results after the event.
  • Fatih: How much of scenario runner CI/CD is created?
    • Tier IV already has the working CI/CD pipeline internally. They will share the updates to related working groups next week.

Working group report: Hardware (10 min)

  • CJ: There is a need to revisit the scope and goal of the hardware working group. CJ will facilitate the meeting between Joe, Rod, Mitsudome, CJ, Larry, Matt, Armagan
    • Yang: I would like to be updated with the discussion
  • Yang: One of the goals was to provide a list of supported sensors. Whatā€™s the progress on publication on the document?
    • The topic has fallen off the agenda.
    • This should be raised up again

Working group report: Simulation (10 min)

Working group report: Racing (10 min)

  • Main Achievements:
    • Providing summary of interesting material to stat with the topic ā†’ Attract people to join the working group
    • Development of new SVL racetrack on the 1:10 scale (Red Bull Ring Austria)
  • Current Hurdles and Work in Progress:
    • Bringing Autoware.Auto alive on the F1TENTH vehicle. This work is progressing a little bit slowly through the porting of Autoware.Auto to the F1TENTH vehicle: Missing interface, missing correct VESC message exchange. And will take additional time ā†’ Solving this by assigning new software developers
    • Developing new algorithms (perception, planning and control) for the F1TENTH vehicle. E.g. the current localization algorithm in autoware.auto is not capable of dealing with 2D lidar information. When autoware.auto is running on the vehicle we can try different 2D localization techniques (e.g. GoogleCartographer) or Indoor GPS Navigation for the vehicle
  • Future to Dos:

Discussion:

  • CJ: We would not be able to meet the milestone for setting F1Tenth vehicle for training starting from January. Fallback option is to use simulator for the training course.
  • Fatih: If someone needs to use Autoware for 1/10 scale vehicle racing, we need to spend at least half a year.
    • CJ: We are considering using 3rd party software to integrate with Autoware.Auto to fill in the missing gaps.

Working group report: Open AD Kit (10 min)

Discussion:

  • Shinpei: How is the discussion with MIH going?
    • CJ: we have proposed to create a working group to MIH TSC
    • We are still working to do a kick-off meeting

Proposal for Autoware Reference Design (Joint proposal)

Reference Design 102021 TSC Meeting.pdf (381.7 KB)

  • We are expecting to have outputs from each TSC working group.
  • We should have comprehensive documentation on requirements of autonomous driving stack

Autoware.Core/Universe Update (Tier IV)

Discussion:

  • How is this going to proceed? Is it going to be led by Open AD Kit?
    • Tanzawa: I will provide a basic draft first and bring it to the Open AD Kit. Further discussion can be decided there.
  • Fatih: Is Autoware.Universe going to be in a separate repository or is it going to be a branch of Autoware.Core?
    • Ryohsuke: Current plan is that we will set up a separate repository. The diagram in this presentation is referring to the documentation
  • Filipe: Do you have any additional details?
  • Yang: I surely support the idea, but the biggest concern is timeline of the execution
    • Tanzawa: We will be providing more than 10 engineers for execution
    • Yang: Iā€™m not worried about resource, but we would like to have publicly available document about execution plan
  • Shinpei: Are there any other packages that could go into the Universe?
    • Ryohsuke: I still havenā€™t received any proposal yet.
    • Shinpei: Maybe Adastec and ITRI could provide their modules used for Bus ODD
    • Michael: I have to check with my coll
    • Shinpei: When you create Core packages, I would like to make sure that functional safety is not ignored
    • Fatih: How are the packages maintained in Universe? How does the modules in Universe communicate each other?
      • The guidelines about the maintenance process of packages was shared by Jilada in the past, but we can discuss in more detail in a separate call
      • Action item: Mitsudome to set up a call among software architects to review on Core/Universe proposal.
    • Stephen: AutoCore is supportive of the idea of both Core/Universe and Reference Design.
1 Like