Autoware Foundation premium member contributions

This topic is to gather the contributions that Premium members of the Autoware Foundation plan to make to the development of Autoware and other Autoware Foundation projects.

This is a public topic. We welcome comments from the Autoware community about what they would like to see developed for Autoware, variations on proposed contributions, etc.

I am replying to above action item. Apex.AI will contribute:

  1. https://gitlab.com/AutowareAuto/AutowareAuto/issues?scope=all&utf8=✓&state=opened&assignee_username[]=esteve
  2. we will also find a parking lot for testing in California

Regarding StreetDrone, at the moment the contributions will be on the following:

  1. Vehicle Interface
    Based on #62 and #1541. An amount of that work has been published here.
  2. Vehicle Modeling (URDF, Gazebo)
    An amount of work has been published here.

Also, I agree on a single use case each time with scalable complexity (Willow Garage style approach) although I understand other companies could have different priorities.

2 Likes

Parkopedia will review Lanelets and other map formats considered by Autoware and provide feedback at the next TSC. If all goes well Parkopedia will begin the work of integrating Lanelets into Autoware, writing the code to convert our map representation into OSM-XML and to supply a small number of maps of parking facilities for other members to use within both the simulator and their test vehicles.

I also intend to contribute the safety case that we have developed with our project partners the Transport Systems Catapult (now the Connected Places Catapult). Some initial documents should be available within the next few days.

2 Likes

I’m really looking forward to seeing your safety case.

When you release the documents, please make it clear what the IP restrictions are.

AutoCore.ai will contribute follow:

  1. ADCU design as autoware.io product supports autoware.ai/auto on heterogeneous computing platform (commercial hardware products)
  2. Open source vector map editor for autoware internal map format (with LGPL license)
1 Like

Linaro:

  1. Autoware.ai CI, code quality, test coverage improvement
  2. Arm based SoC Autoware enablement through Autoware.io project
1 Like

Tier IV will contribute to porting Autoware.AI algorithms to ROS 2 for Autoware.Auto.

Here is a high-level overview of the working items Arm is looking into (or are planning to) at the moment:

ArmNN integration [1]

  • Core (Neon/SVE), GPU and NPU

Optimizations (e.g. Neon, SVE2 [2])

  • 3rd party libraries
  • Computing nodes
  • Flow parallelization for multi-core systems

Yocto-Based Base Platform

  • Reference stack with virtualization support
  • Distro with Autoware dependencies
  • Default machine based on Armv8-A Base FVP [3]

Arm Compiler6 support

  • This is a LLVM based compiler
  • Safety certifications [4]

Launcher/UI

  • Improvements to the launcher allowing easier configuration of nodes executing on multiple systems. This is to provide a better separation of nodes running on the “vehicle” system and nodes running on the “developer” system.

[1] https://developer.arm.com/ip-products/processors/machine-learning/arm-nn
[2] https://community.arm.com/developer/ip-products/processors/b/processors-ip-blog/posts/new-technologies-for-the-arm-a-profile-architecture
[3] https://developer.arm.com/tools-and-software/simulation-models/fixed-virtual-platforms
[4] https://developer.arm.com/tools-and-software/embedded/arm-compiler/safety

Note: They are not in any priority order.

2 Likes

TRI-AD

  • General engineering resources (full time engineer)
  • Infrastructure setup / maintenance / support (CI / CD, etc.)
  • Data processing pipelines (design and potentially future implementation)
1 Like

Any update on when you will be able to provide this information? I need to start thinking about the safety plan for Autoware.Auto milestones soon.

I’ll check with CPC (was TSC) what the status is of the documents. They were undergoing internal review last time I checked.

1 Like

LGE

  • AD Simulator (as open source project and free binary version) (https://lgsvlsimulator.com)
  • 3D maps for simulators (currently, San Francisco, Akihabara (Tokyo), Shalun (Taiwan), and so on).
  • Vehicle models for ego and non-ego vehicles
  • Plan to deliver a new 3D environment and a vehicle model (Lexus) for Autoware TSC members in June.
3 Likes

eSOL

  • Build a working demo of Autoware.Auto on eMCOS(*) jointly with Apex.ai, TierIV and other contributors.

*eMCOS is our commercial scalable RTOS which is POSIX compliant and is to be certified for ISO-26262 ASIL-D.
https://www.esol.com/embedded/emcos.html

1 Like

AutonomouStuff has (and will continue) to provide:

  • In-vehicle testing and fixes from the Open Autonomy Pilot program
  • Improvements to CI
  • Feedback and advice on the infrastructure and future direction
  • Maintenance of some drivers required by the project (Velodyne, Novatel, etc.)
  • Continued development and maintenance of the AS SSC Interface
  • Continued support regarding multiple vector-based mapping formats

AutonomouStuff will soon provide:

  • A new, more generic, TensorFlow-based traffic light classifier (post-1.12)
  • A GNSS/INS-only localization solution
1 Like

Kalray

Contribute Acceleration capabilities for Reference Platform:

  • PCIe boards with MPPA®2 Generation, MPPA®3 upcoming
  • Accelerator daughter board modules (PCIe)
    Reference Programming Software

CNN Inference Optimizer

  • CNN Inference Runtime Engine (additional algo support and growing)
  • Computer Vision OpenCV
  • OpenCL 1.2 runtime

Hardware Integration

  • Support for Intel and ARM hosts

Software Integration

  • Upcoming Additional Use Cases support (Companion, StandAlone) to simplify System integration
  • Linux and Autoware Member RTOS supported