Technical Steering Committee (TSC) Meeting #21 (2020/08/19) Minutes

2020/08/19 22:00 UTC

Attendees

  • Geoffrey Biggs (TSC chair)
  • Hiroshi Igata (Tier IV)
  • David Van Geyn (AutonomouStuff)
  • Jaime Martin Losa (eProsima)
  • Adam Dabrowski (Robotec.AI)
  • Joe Speed (ADLINK)
  • Stefano Longo (Embotech)
  • Ryohsuke Mitsudome (Tier IV)
  • Larry Dimitrievski (Tata Consultancy Services)
  • Josh Whitley (AWF)
  • Christopher Ho (Apex.AI)
  • Tsutomu Otake (Macnica)
  • Tony Lee (Velodyne)
  • Jeff Johnson (Mapless AI)
  • Akihiko Tsukuda (eSOL)
  • Shinpei Kato (AWF)
  • Nobuyoshi Nakajima (ITD Lab)
  • Ali Ors (NXP)
  • Seonman Kim (LGE)
  • Daisuke Tanaka (AWF)
  • Yang Zhang (Linaro)

Minutes: Geoffrey Biggs

Action items

  • Add documentation on how to change the DDS implementation used
    • Josh Whitley
  • Follow up on acknowledgements being given in Project Aslan
    • AWF board/Josh Whitley
  • Change the default DDS implementation used by Autoware.Auto to Eclipse Cyclone DDS
    • Josh Whitley

Opening remarks and new member introductions (Board)

New members:

  • eProsima
  • Tata Consultancy Services (TCS)

Confirmation of previous minutes

Minutes approved.

Action items review

  • Hold a vote by email on changing the default DDS for Autoware.Auto from Fast RTPS to Cyclone DDS
    • Geoffrey Biggs
    • Action taken: Vote was held
  • Follow up on acknowledgements being given in Project Aslan
    • Josh Whitley
    • Action taken: This is now waiting on the exact text for Project Aslan to use from the Board.

New Autoware Foundation project: Autocore Sim

Autoware Foundation social media accounts

The Autoware Foundation now has social media accounts. Please follow these accounts and pass them on to your friends so we can increase the visibility of the AWF.

FTE technical contact details

Several companies who have committed to provide engineering time have not yet sent in the contact details for their technical contact. These companies will be asked directly to provide the contact details. If you know you have not provided them, please do so now.

Vote on changing the default DDS to Eclipse Cyclone DDS

Proposal:

  • That the default DDS implementation used for Autoware.Auto be changed from Fast RTPS to Eclipse Cyclone DDS.

For: 7

Against: 2

Abstain: 14

Proposal is accepted.

Working group report: Autonomy Software

The ASWG finally closed out Milestone 2 this month and the tag ms2 is now on the repository with a working demonstration launch file in the autoware_auto_avp_demo package. Milestone 3 is moving along. A service interface for querying map data was defined and implemented. All but one of the major features in the Planning and Navigation architecture are nearing review stage. Parkopedia received map data from our mapping provider and will be providing a downsampled point cloud map and Lanelet2-format map very soon. They are also nearly finished with the first implementation of the Behavior Planner. However, we have many items which are in need of developer contributions. Please see this link for a list of issues in need of attention!

Working group report: Autonomy Hardware

The traction on HW Working Group has been slow. Despite few progresses, we are not in a position to make this group fully valuable as it stands now.

For these reasons we would like to switch gears and actually propose the following:

1- Promote existing Reference Hardware from AWF member, namely Autocore with the PCU. Establish link on official website autoware.org

Objective: position Autoware as a solution provider (hardware + software)

Interest: Raise up interest on hardware availability and question

Timing: between now and end end of October

Actions: Autocore to prepare information for publication / Autoware TSC to review

Risk: availability and support (Autocore)

2- HW Working Group to receive ODD (AVP definition and Cargo Delivery definition) and compare feasibility on Reference HW. Establish needs for coming hardware: From ODD and associated cases for needs (interfaces, compute capabilities, demand prod, …)

This activity is aligned with what has been proposed earlier this year. The facts to (1)link this to a more formal ODD and (2) an initial implementation of AVP will help progressing.

Reminder: this activity only rely on AWF member to participate.

Action one will raise interest and help on getting more visibility and hopefully moire attendance to calls

Working group report: Operational Design Domain

Continuing weekly meetings which cycles through the time zones of the regular participants.

Have setup the working group wiki page.

Reached consensus on the cargo delivery with indoor travel ODD.

Initiated discussion on the use cases supporting the ODD starting from the Tier IV-Yamaha cargo delivery use cases.

Discussing whether we need to simplify the use case list. (ex. Eliminate multi lane cases and remotely-triggered cases.)

Started discussion on the machine readable scenario description format.

Ridecell, LGE, and Tier IV made proposals on scenario format.

Building a comparison table to make decision. Evaluation points include open standard, productivity, capability to handle the parameter sweeps and the complex scenarios, etc.

Choosing another ODD WG co-lead

The ODD WG currently only has one leader, whereas two are required. A volunteer is necessary for the second leader.

LG will consider if they can co-lead.

FTE activity reports review

FTE activity reports 2020_08_19.pdf (69.8 KB)

Autoware.AI 1.15 release dates

This will be the final minor release of Autoware.AI before it goes into maintenance mode. We need to make a release by the end of 2020.

Recommendations from Josh:

  • Close PR submission at October 1st
  • PR integration and testing until end of November
  • Release on December 1st.

AutonomouStuff would like a little more time because they have a large number of complex PRs to send in.

Chosen dates:

  • Close PR submission at October 23rd
  • PR integration and testing until December 4th
  • Release on December 11th

Long-term ODD roadmap

  • Tier IV wishes to begin planning the ODDs to work on after the current next ODD (cargo delivery in a closed environment) is completed.
  • After cargo delivery, Tier IV proposes doing active campus (controlled but open area) passenger shuttles.
  • Following that, Tier IV proposes doing autonomous taxis in urban areas, limited to daytime. Tier IV believes this should be the long-term goal for Autoware.

Discussion

  • AS: Is the intention to go straight to autonomous taxis or to do something in between?
    • Tier IV: The idea is to have intermediate ODDs before we go to autonomous taxis. Tier IV’s proposal is to do campus navigation in between cargo delivery and autonomous taxis, but we could perhaps have an intermediate step between campus navigation and autonomous taxis, such as autonomous taxis in a rural environment.
  • AWF: How should we implement large feature sets that we want in the Autoware.Auto codebase but which are not required by a specific ODD, such as extended sensor support? Where in our development processes do we do the work to support these? Should we have a non-ODD-related development cycle?
    • Tier IV: We can add non-ODD-required features to the codebase, and create launch files to start the ODD-related applications using alternative sensors, etc.
  • Is “autonomous taxis” a suitable long-term goal?
    • ADLINKL: It should not be the only long-term goal. Things like autonomous trucks are also important.
    • Board: We need to have clear long-term goals and this is a good one to encourage people to join the AWF and use Autoware.
  • The board prefers that long-term goals be set by the Strategy Planning Committee.

Correction: The Autoware Foundation’s Twitter account is now https://twitter.com/AutowareFdn.