Technical Steering Committee (TSC) Meeting #36 (2021/11/17) Minutes

Autoware
1

2021/11/17 22:00 UTC

Attendees

  • Ryohsuke Mitsudome (TSC Chair)
  • Shinpei Kato (AWF)
  • Daisuke Tanaka (AWF)
  • Christian John (SPC)
  • Bonolo Mathibela (AWF)
  • Xinyu Wang (AWF)
  • M. Fatih Cırıt (AWF)
  • Hiroshi Igata (ODD WG)
  • Wojciech Jaworski (Software WG)
  • Kasper Ornstein Mecklenburg (Arm)
  • Stephen Li (AutoCore)
  • David Van Guyen (AutonomouStuff)
  • Rahul Mangharam (F1 Tenth Foundation)
  • Michael Li (ITRI)
  • Mert Colak (Leo Drive)
  • Lokesh Kumar Goel (LGE)
  • Adam Dabrowski (Robotec.ai)
  • Naoyori TANZAWA (Tier IV)

Minutes:

Action items

Confirmation of previous minutes

Board Passdowns (5min)

  • Ryohsuke to on-board Haoru
  • Are we still hiring one more Software Developer?
    • Shinpei: We can keep the career page open to look for new talents
    • Daisuke: We are currently closing the page right now until situations the other positions are clear
    • Shinpei: We should talk this further in the BoD meeting, but we want to keep position open

New member update:

  • LeoDrive representative: Mert Colak
  • LG representative: Lokesh Kumar Goel
  • ADLINK representative: Henry
    • Ryohsuke to check with him if he is receiving the invite
  • Software architect: Esteve, Xinyu, Fatih

Working group report: Autonomy Software (10 min)

Working group report: Operational Design Domain (10 min)

  • The demonstration and walkthrough of the (scenario simulation) CI/CD pipeline which was proposed to the Autoware Foundation via the simulation working group

  • The ODD working group members have been encouraged to register for this tool so that they can contribute to the creation of the scenarios for the future main projects of the Foundation.

  • The cargo delivery scenarios which had been created by the ODD WG using the scenario testing framework are now copied to the scenario storage of the new CI/CD pipeline and some of them have been brushed up as the basic set for the scenario testing.

  • LGE worked diligently to connect the SVL simulator to the scenario runner using ZeroMQ interface, and you will be able to run the scenarios created by the ODD WG with the SVL simulator as well.

  • Started the preliminary discussion of the public road bus service scenarios based on the information provided by ITRI.

  • Scenario Tool Summary has been added to the ODD WG Wiki page for clarification of the usage/purpose of “the scenario testing framework” and “the CI/CD pipeline”. The former is the local scenario simulation for any Autoware.Auto projects and the latter is the cloud-based DevOps environment for the primary project of the Autoware Foundation including cargo delivery.

  • How are we proceeding with Bus ODD?

    • Igata: we are going over use cases in bus scenario

Working group report: Hardware (10 min)

Shinpei:

  • Has Joe already left all AWF activities? E.g., SPC
    • Joe was sharing information in SNS. Is anyone taking over his role&responsibility
    • Since Joe was active in different AWF activities, we should contact Joe to make sure that all items are handed-over to someone else
    • We can make some time in the next call to say a word to TSC as well
    • Action Item: Ryohsuke to contact Henry and Joe for the next call

Working group report: Simulation (10 min)

Working group report: Racing (10 min)

  • Autoware.Auto porting to the F1TENTH Vehicle:

    • Currently this is the main effort we are doing and this is taking the most time
    • We have finished the VESC ROS2 driver which is no enabling a full access to the F1TENTH vehicle (actuation and sensors) and is completely running on ROS2
    • We have sent an F1TENTH car to Japan to Mitsudome so he can follow up with the work from Josh Witley. He will finish the Autoware.Auto porting and bring it to run on the car
    • Currently we are missing a 2D localization technique that is integrated in autoware.auto but is needed for the 2D lidar on the F1TENTH vehicle. We have previous algorithm from F1TENTH available (google cartographer, particle filter) but their usage in ROS2 is limited. Together with Xinyu Wang we are exploring the capabilities of those algorithms and will then implement them in the Autoware.Auto pipeline. The main algorithm will be probably google cartopgrapher because this one was developed for the usage of 2D lidar and is ported to ROS2 already.
    • Planning to have a hackathon on Dec 10 at Circuit Launch with the goal of reliably demonstrating waypoint recording and playback. Need to address the localization topic to get this to work reliably due “noisy” environment (people around track), track setup (low barriers, moving cones, etc.) and car vibration (e.g. z-axis movement introduces noise into recorded 2-D lidar data)

  • Simulator Development:

    • The SVL Simulator F1TENTH vehicle has now two different Sensor setups. One is the original sensor setup from the F1TENTH car (2D Lidar, Intel Realsense) and the other sensup is called BIG CAR (3D Lidar, GPS, Camera) to enable the sensor behavior of a real vehicle. The latter one can be used to run Autoware.Auto with the F1TENTH vehicle to gain more experience of the vehicle behavior and the usage of Autoware.Auto for interested developers
    • We created a new F1TENTH racetrack which is modelled after a real racetrack (Red Bull Ring Austria) and uploaded this model to the SVL Simulator. https://www.youtube.com/watch?v=aO72j5dZEg0&t=1s - the possibilities with this racetrack are endless (sponsoring ads, night races) and we can scale the track up to run the Indy cars on it, too.
    • We are setting up a Youtube Tutorial called “How to race in the SVL simulator” that covers all aspects of running with the F1TENTH car in the SVL simulator and have some basic race algorithms (pure pursuit, follow the gap, graph based planner)

  • Upcoming Events:

This is a list of all upcoming events where F1TENTH/ Autoware is involved. All of the Events will be in Person but both CPS week and F1TENTH will be virtual too. We are exploring currently the usage of the SVL simulator for the race → Perhaps there is some feedback from the AI Challenge and how we can handle that effectifly?

Working group report: Open AD Kit (10 min)

  • Working well towards a 1.0 release with Bonolo hosting a meeting coordinating activities
  • SOAFEE reference implementation EWAOL now integrates an orchestrator (k3s), making it a viable OpenAD kit base architecture
  • EWAOL integrated onto the ADLink AVA platform
    • We also have EWAOL working on other platforms (RPi4, NVidia Xavier, IMX8, …) but there are not officially supported
    • EWAOL is also functional on AWS Graviton to enable OpenAD Kit to be deployed unmodified in the cloud
    • These come together to give us a broad ecosystem reach for OpenAD Kit
  • Conversations starting with Autonomy SW WG to understand containerization requirements for microservices based deployment of Autoware
  • Deployment of the OpenAD Kit stalled for the time being whilst we understand the roadmap for AWS Storefront white-labeling

Joint Proposal on updating Autoware.Auto

Autoware.Auto upgrade 112021 TSC Meeting.pdf (1011.0 KB)

Joint proposal from Arm/AutoCore/ITRI/LeoDrive/Robotec/Tier IV

  • Support ROS2 Galactic
  • Create Autoware.Core newly
  • Adopt GitHub
  • Define interface for more scalability

Q&A:

  • Shinpei: any downside of using Github?
    • Bonolo:
      • we loose git history
      • there’s difference in Project/Milestone
      • We still need to be careful
  • Should we make a vote on this?
    • Shinpei: We had a vote last year so we don’ t have to do it again
    • Since the situation might have changed since then, it’s worth doing it again
    • Suggestion from Ryohsuke: We create a prototype of Core/Universe repositories in Github and give a report in the next call to make a final decision about the transition
  • https://docs.google.com/document/d/18hSPEp8oZl_YBNMdwxPl1sxaJaog9526eyEEB8rAY_Y/edit#heading=h.gjdgxs
  • How do we proceed with defining the interface?
    • Open AD Kit is working on creating a reference design for Autoware
    • Open AD Kit WG will contact relevant WGs to fill in the design

Formation of Safety Working Group

Safety WG 111721 TSC Meeting.pdf (741.7 KB)

  • Ryohsuke: One concern that I have is the scope and deliverables
    • Safety usually requires a very large amount of effort to do safety analysis and defining safety requirements.
    • CJ: we will continue working on defining the scope in SPC and raise this topic again in the next TSC meeting to finalize
    • Stephen: The relationship between MIH WG should be made clear. We don’t want to have a duplicate WG.
    • Naoyori: we need to define what is safety
2 Likes