Technical Steering Committee (TSC) Meeting #31 (2021/06/16) Minutes

Autoware
1

2021/06/16 15:00 UTC

Attendees

  • Ryohsuke Mitsudome (TSC Chair)

  • Shinpei Kato (AWF)

  • Yang Zhang (AWF)

  • Bobby Hambrick (AWF)

  • Jim Liu (AWF)

  • Girish Shirasat (AWF)

  • Daisuke Tanaka (AWF)

  • Christian John (SPC)

  • Josh Whitly (AWF)

  • Bonolo Mathibela (AWF)

  • Hiroshi Igata (ODD WG)

  • Kerem Par (Adastec)

  • Joe Speed (ADLINK)

  • Filipe Rinaldi (Arm)

  • Cheng Chen (AutoCore)

  • David Van Guyen (AutonomouStuff)

  • Jaime Martin Losa (eProsima)

  • Akihiko Tsukuda (eSOL)

  • Rahul Mangharam (F1 Tenth Foundation)

  • Michael Li (ITRI)

  • M. Fatih Cırıt (Leo Drive)

  • Seonman Kim (LGE)

  • Tsutomu Otake (Macnica)

  • Pascal Lecuyot (Navya)

  • Adam Dabrowski (Robotec.ai)

  • Jilada Eccleston (Tier IV)

  • Yao Zhai (Ziiko)

  • Luxiang Zhou (guest speaker)

  • Shark Liu (guest speaker)

  • Abraham Monrroy (guest speaker)

  • David Wong (guest speaker)

Board Passdowns

  • SPC:
    • Hiring proposal for 1 or 2 developers to support Josh/Bonolo
    • Website update in progress
    • Cross-alliance with AVCC (Autonomous Vehicle Compute Consortium)
  • OC:
    • Discussion on setting a grace period for a member company is exempt from membership obligations
  • Membership updates
    • Industry Level: Polyexplore
    • Premium: Leodrive, ITR(upgrade from academic & non-profit members)

Confirmation of previous minutes

Working group report: Autonomy Software

  • Temporarily assigned Frederik (Apex) to run one of the three time slots.
  • Perception stack development done by Apex and LeoDrive
    • First draft for tracking module and prediction module were made
  • Architecture Design for cargo is done for all major components
  • Localization module was updated to support GNSS
  • Integrated Merge-Bot to automatically merge a MR once approved by a reviewer
  • Improvements going on for CI pipeline to speed up the process
  • Planning architecture was decided, and some modules started to be ported from Tier IV’s reference architecture
  • We are expecting map data of Seattle Airport from PolyExplore next week

Discussions & Questions:

  • Joe: Should Nav2 be involved?
    • Josh: We should wait for Cargo ODD to be done.
  • Shinpei: How was the perception algorithm decided?
    • Tier IV and Apex had a separate call to discuss this.
  • Shinpei: Do you have any trouble doing project management?
    • Josh: We just asked Fred to lead the call.
    • How is engineer management done?
      • Currently all issues are tracked by milestones.
  • Filipe: Are we moving to c++17?
    • Josh: We are aiming to support Galactic, and we should be supporting c++17 as part of this activity
    • Filipe: We are planning to donate a functional safety compiler, but this may become a blocker.
    • Action Item: Josh + Filipe to have a separate call to decide how we should go with C++17.

Working group report: Operational Design Domain

SeaTac airport has been chosen as the demonstration site and Zing Robotics kindly provided the video recording of the entire path of the human operated cargo vehicle.

The scenes of interest have been identified by Alex (Nagoya University), and ODD WG discussed how we can potentially handle those scenes. As the Autoware foundation demonstration course and the detailed content has not been defined yet, Alex brought the list of the scenes to the Autonomy Software working group for discussion.

Once the scenes (use cases) to be addressed in terms of simulation scenarios are identified, ODD working group is to write additional scenarios both in natural language and the machine readable format.

Coming milestones:

  • Identify additional scenarios for cargo Delivery ODD by E/June.
  • Create natural language & machine readable scenarios by E/July.

Steve and I have been participating in Racing WG and working on the ODD definition of racing in comparison with other ODDs (Cargo and AVP).

Working group report: Hardware

Working group report: Simulation

  • Chair and Co-chair have been decided
    • Chair : Wojciech Jaworski from Robotec.ai
    • Co-chair : Joseph Fernando from ARM
  • Work organization:
    • Meeting of Simulation Working group will be organized every 2 weeks.
    • Workspace for daily cooperation will be arranged until the end of this week:
      • Landing Page on Gitlab
      • Discussion tasks about major topics on Gitlab
      • ROS discourse thread
  • Technical discussion:
    • Sim WG will closely cooperate with ODD group in terms of Scenario generation and requirements
    • Initial idea is using TierIV ScenarioRunner as starting point
    • Discussion about the architecture of the entire Simulation Pipeline will be continued on Gitlab and biweekly meetings
    • The most important topic for now is to decide Scenario Format enabling randomization/variation of parameters
    • Together with the ODD group we should define metrics used for scenario evaluation.
    • Definition of correctness of scenario should be included in testing procedure together with scenarios definition (ODD group should be responsible)
  • Minutes:

Working group report: Racing

  • Working with ODD WG leads (Igata-san/Steve) on high level definition of the Racing ODD.
  • Planning for Autoware.Auto hackathon event August 12-14 targeting to complete basic integration of Autoware.Auto on the F1TENTH hardware platform. Hackathon will be held in the SF Bay Area and coincide with an existing DIY Robocars event which attracts many AD developers/students. Milestone is key to achieving the goal of enabling F1TENTH with Autoware.Auto for the start of University programs in Sept.
    • Is this a replacement of Hackathon that was originally planned for Cargo?
      • → Yes. We can expect more participation in the hackathon.
  • Team has been supporting the IAC Base Software WG to bring up Autoware.Auto on the Clemson University built racecar.

IAC update Jun 16th for Autoware TSC.pdf (2.3 MB)

Generic Lidar Device Driver Proposal (Tier IV & Nagoya Univ.)

Presenter: David + Abraham
LidarDriver_TierIV.pdf (1.4 MB)

Comments & Questions:

  • Fatih: There should be generic way of using GNSS time as timestamp. We should also adding timestamp for each point.
    • David: It is true that different hardwares configure timestamp differently
  • Xingang Liu: How is the distortion of point cloud due to ego vehicle motion?
    • David: There is a question whether that should be included in the driver or it should be as post-processing.
    • Josh: In Autoware.Auto, it is designed to do it in post-processing.
  • Xingang: There could be shadows(noise) in multi-lidar situations.
    • David: We are thinking of doing filtering in post-processing, but we are open to new ideas about the pipeline.
  • Ryohsuke: Is it done in ROS1 or ROS2
    • First implementation would be in ROS2, but we will have ROS wrapper to support both ROS1 and ROS2
  • Josh: Is implementation in Autoware.Auto considered?
    • Abraham: yes
  • Shinpei: What’s the next step?
    • This is a proposal on a generic lidar device driver for Autoware.Auto. Wouldn’t this have a significant impact on the current code base?
    • This can be a basis for Autoware.IO.
  • Josh, David, Ryohsuke could have a separate call to move forward.

Introduction on SDV Project (AutoCore)

In the past half-year, AutoCore is developing a project named SDV based on V2X and K8s. It can deploy software stack to the edge by k8s and monitor and set targets of multi-vehicles from cloud. Currently, it’s integrated with Polit.Auto’s ROS2 version. We think it should be useful for Cargo ODD. We are willing to contribute this project to AWF, and start integrating with Autoware.auto for Cargo ODD.

  • Josh: Regarding the interface, how was API decided? What kind of information is passed between edge and cloud? Autoware.Auto does not have V2X interface defined. This can be a starting point to initiate this feature.
    • Are there design documents on the interface?
      • They are in Chinese and still immature.
      • V2V and V2X should be separately defined as well.
  • How are we moving forward?
    • OpenADKit focus on container and CI/CD and not on V2X.
    • We might need a separate working group to discuss this topic.

Core/Universe Proposal Update (Tier IV)

CoreUniverseArchitectureProposal_2021_06_17.pdf (862.8 KB)

Next Step:

  • Ryohsuke will send out a vote whether to support this concept.

FTE activity reports review

FTE activity reports 2021_06_16.pdf (65.0 KB)