Technical Steering Committee (TSC) Meeting #34 (2021/09/15) Minutes

Autoware
1

2021/09/15 22:00 UTC

Attendees

  • Ryohsuke Mitsudome (TSC Chair)
  • Shinpei Kato (AWF)
  • Daisuke Tanaka (AWF)
  • Christian John (SPC)
  • Bonolo Mathibela (AWF)
  • Hiroshi Igata (ODD WG)
  • Joe Speed (ADLINK)
  • Stephen Li (AutoCore)
  • Girish Shirasat(Arm)
  • Michael Li (ITRI)
  • Larry Dimitrievski (TCS)
  • Jilada Eccleston (Tier IV)
  • Naoyori Tanzawa (Tier IV)
  • Seonman Kim (LGE)

Minutes:

Action items

Board Passdowns (5min)

  • Membership updates:
    • Leaving: ITD Lab, RTI, FLIR Mandii TBD
    • Joining: Gaia(Industry level), Agile X(Industry level, application on-going), Civil Maps(Industry level, application on-going)

Call for new Software Architect (Operation Committee) (5min)

  • plan to address Josh’s departure thru call for participation from member companies to commit ~3-4FTE thru additional incentive mechanism (there would not be counted as regular FTEs)
    • Incentive mechanism includes AWF paying for additional FTE commitment (outsourcing of development)
  • Official call for the Software Architect will be shared via email.

Change in the representative of Tier IV

Jilada → Naoyori Tanzawa

Confirmation of previous minutes

https://discourse.ros.org/t/autoware-technical-steering-committee-tsc-meeting-33-2021-08-18-minutes/22257/2

Action items review

Working group report: Autonomy Software (10 min)

  • MS 5 closed.
  • We’ve had significant improvements on controllers and tracker. The author of the original controller we made for AVP is no longer contributing to Autoware and no one was able to tune parameters, but now we have a simpler controller which is easier for tuning. Tracker can now fuse detected objects from lidars and cameras. It has been tested with LG SVL, but still needs testing with real data.
  • Some of the ongoing items include implementation of the planning stack, system monitoring modules, data collection for training perception and preparation of the launch files for cargo delivery.
  • Preparation for the demo is ongoing. Zing’s vehicle implementation is tracked in this issue. Weekly meetings are held separately with Zing to track progress. The deadline for some of the action items has changed, but the overall deadline stays on 12th November. Adastec also volunteered to do some in vehicle testing of some of the new features preceding the demo for debugging purposes.
  • Other items:
    • .Core/Universe idea shared to the working group members
    • Request from Racing working group to do localization with 2D lidars and stereo cameras.

Working group report: Operational Design Domain (10 min)

  • The forty-three cargo delivery scenarios plus additional four scenarios derived from the Seattle Tacoma airport video have been checked, revised and confirmed to be compatible with the latest version of the scenario testing framework. Some of the scenarios may better be rewritten using the actual map data from Seattle Tacoma airport.

  • Have been discussing how to obtain the map data of Seattle Tacoma airport both indoor and outdoor firstly for the simulation use. If the point cloud data is obtained, we can create the lanelet2 map by using the Vector Map Builder (Autoware tools), regardless of the quality of the point cloud map. This is viable for the scenario simulation in the scenario testing framework.

  • Have completed the ODD definition table of the bus service based on the information provided by ITRI. This is a reference for the next ODD discussion.

  • Tool update: LGE is working on the ZeroMQ-Python interface which enables SVL simulator to run the same scenarios created with the scenario testing framework in the ODD working group so far.

Working group report: Hardware (10 min)

  • How would the LeoDrives document be shared?
    • Still needs discussion in SPC.

Working group report: Simulation (10 min)

Working group report: Racing (10 min)

  • Working group attendance and meetings are progressing well
  • Overall refinement and definition of WG goals and and activities on the gitlab page: Racing working group · Wiki · Autoware Foundation / Autoware Foundation · GitLab
  • In person meetup of members in California at the DIY robocar event.
  • Hackathon for further development of the software for the F1TENTH vehicle at the DIY robocar event in California
  • Main Achievements:
    • F1TENTH vehicle as Standard Platform: We provide the Bill of Material and an in-depth build manual (slides, videos) to set up a small-scale autonomous racing vehicle. This vehicle is the standard testing platform for Autoware.Auto, is allowed to race in racing events and is the baseline for further developments.
    • Racing ODD Definition: This file includes a high level definition of what we need to take care about when it comes to autonomous racing with Autoware.Auto. We define the possible physical infrastructures, operation limits, possible objects to detect etc. This is compliant to other Autoware.Auto ODDs and creates a machine readable file for the automatic generation of test scenarios.
    • Racing in SVL Simulator: The SVL simulator enables the simulation of racing vehicles (F1TENTH, IAC racecar) as well es different racetrack environments (Indianapolis Motor Speedway, Indianapolis Motor Speedway 1:10 scale, DiY Robocar Circuit Launch)
  • Current Hurdles and Work in Progress:
    • Bringing Autoware.Auto alive on the F1TENTH vehicle. This work is more complicated than thought in the beginning (missing ROS2 drivers, creating branch for the F1TENTH vehicle only, creation of new vehicle interface for the F1TENTH vehicle) and will take additional time
    • 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
  • Future to Dos:

Working group report: Open AD Kit (10 min)

Initial meetings have been progressing well, with an agreement on the Vision and Mission for the project.

Tier4 have delivered an internal release of the 0.5 Open AD Kit architecture which starts the migration towards a Software Defined implementation.

The public release of SOAFEE took place on the 15th September, and internally Arm has the SOAFEE architecture running Autoware.Auto ROS 2 Cyclone DDS. This is the first step towards the deployable OpenAD Kit vision.

We have discussed a roadmap beyond the 1.0 release, and this is to be presented to the SW architecture group for inclusion in their planning.

All meeting minutes and presentations are available on the Autoware Wiki.

Updates on Cargo ODD Demo

SPC and Board agreed to run a Cargo Delivery demo using eve autonomy’s vehicle and warehouse. Planned to be done in mid-October to mid-November timeframe.
EveAutonomyCargoDemo.pdf (638.9 KB)

Follow up on next ODD Planning (CJ)

ODD Roadmap Decision 091521 TSC Meeting.pptx.pdf (1.4 MB)

Would like to have a vote on the next ODD. Three options as a candidate:

  • Shuttle/Bus operating on restricted private roads (e.g. industrial complex or business campus) at low speeds and exposed to limited NPCs
  • Shuttle/Bus operating on public roads at low-mid speeds and greater exposure to NPCs (frequency, ERV, etc.) and limitations on fully autonomous maneuvers (e.g. lane change initiated only by safety operator but executed by AD)
  • Shuttle/Bus operating on public roads at low-mid speeds and greater exposure to NPCs (frequency, ERV, etc.) and supporting fully autonomous maneuvers (e.g. AD initiated and executed lane change)

Q&A, comments:

  • Who would be running the demo?
    • ITRI, Adastec, Tier IV would be the candidates to provide platforms.
  • Ryohsuke: I would like to go for public roads (option 2&3) partly due to promotional reasons. With consideration of safety, it would be too ambitious to go for fully autonomous driving on public roads.
  • ITRI: We are already starting public transport testing. It is operating in ROS1, but we can port them to ROS2. If we can contribute all components to Autoware.Auto, we should be able to go for option 3
  • Shinpei: option1 would be an incremental update, but considering member companies’ interest in public road use cases, I feel we are capable of accelerating development and aim for public roads.
  • Ryohsuke: Wouldn’t it be risky to say that we are capable of driving on public roads? → Shinpei: We already have many users that are using Autoware on public roads already. We should rather aim to support public road with proper guidelines and limitations

Action Item:

  • Ryohsuke to send out voting emails to TSC members. I will attach today’s meeting recordings for the members who are absent. Some of the comments and concerns that were raised in this meeting should also be included in the notice for reference. CJ can check the draft to finalize the voting notice.

Autoware.Core/Autoware.Universe Update (Tier IV)

Bonolo + Jilada + Tanzawa planned a timeline for Core/Universe Architecture restructure.

  • Is feedback from Josh already reflected? → yes.
  • Did we already agree on going forward with core/universe architecture? → yes, we have already done voting and the proposal was accepted
  • The timeline was designed before EveAutonomy’s demo showed up so dates on the timeline might change, but the structure should stay the same.

SOAFEE + AVA development platform reference ECU platform for Autoware launched today by Arm and ADLINK

https://soafee.io

Some of today’s press about this launch:

FTE activity reports review (10 min)

  • Last month, Bonolo organized members who have no contribution continuously. If they still haven’t shared the report for this month, we will bring it to the Board meeting.
  • There are some members who have contributions to share but haven’t written the report. Ryohsuke will remind them to write them.