AVCC Spring 2024 Virtual Meeting Important Information

When: June 13, 2024 at 9:00 am EDT (1 pm UTC / 6 am PDT)

Where: Online – This event occurred in the past. If you would like access to the recording, please email [email protected].

Agenda

9:00 – 9:10 ET: Intro and Welcome – Elizabeth Kao, Chair of the AVCC Board

9:10-9:55 ET: Guest Speaker  – Brock Walquist, Senior Analyst, S&P Global Mobility

10:00 – 11:10 ET: Working Group Presentations – Highlights/Achievements and Future Work

• • Benchmarking / ABTF – Kasper Mecklenburg
    • AI Benchmarking for Automated and Assisted Driving Systems: The AVCC TR-007 is the third and last technical report on AI benchmarking recommendations for automotive written by the Benchmarks Working Group. It defines AVCC compute scenarios for automotive AI pipelines. These compute scenarios are the building blocks of complex automotive AI pipelines. These pipelines are difficult and can be meaningless to benchmark, so, decomposing them into benchmarkable blocks aligns the automotive industry and makes collaboration more fruitful. In parallel to writing TR-007, AVCC and MLCommons have been collaborating over the past year to make these recommendations into an actual benchmark. This POC will be released in June and is a huge milestone for the AVCC Benchmarks WG.
  • Software Portability / Data Architectures – Rajive Joshi
    • Common Communication Data Architecture for Automated and Assisted Driving Systems: AVCC TR-005 lays the foundation for a common communication data architecture standard, and identifies the need for a common extensible baseline data model to simplify the development and integration of automated and assisted driving systems. This talk will provide an overview of the common communication data architecture. It will explain how interested stakeholders can benefit from, and contribute to the common extensible baseline data model. The data model is aimed at enabling an ecosystem of independently developed reusable software components, streamline software development, reduce software system integration risk, cut software costs, and shorten timelines.
  • Total Cost of Ownership (TCO) – Volker Niemz
    • Evaluating the Total Cost of Ownership for Automated Vehicle Compute: When developing a new compute system for advanced ADAS systems, the system cost is much more than the total hardware cost. AVCC is analyzing the major cost drivers for developing the compute platform with a focus on the software elements that drive the complexity and effort. This talk looks at the problem of porting software to a new compute platform and how AVCC is developing a cost model with the goal of identifying where the industry can benefit in reducing compute costs.
  • AI Cybersecurity – Jonathan Petit
    • Developing Security Performance Profiles for Automated Vehicle AI System: This Summer 2024, the AVCC WG Cybersecurity will kick-off a new study on the security of AI used for Automated Driving. One objective is to develop AI benchmark and dataset to test security. Indeed, AI is a critical component for automated driving, especially in perception tasks. In this talk I will give an overview of the topic and call for participation.

11:10 – 11:55 ET: Rob Dimond, System Architect & Fellow, Arm

11:55 – 12:00 ET: Closing Comments – Stephen Miller, Chair of the Technical Committee

Special announcement! Immediately following this meeting AVCC’s Kasper Mecklenburg will be hosting a webinar with ABTF co-chair, James Goel from MLCommons. Join them to hear more about their big announcement this week on the industry’s first Automotive Benchmark Proof-of-Concept  release.