Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 7 views

    WP.29-194-34e

    Uploaded by

    yosri baati

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    WP.29-194-34e

    Uploaded by

    yosri baati
    7 views4 pages

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    7 views4 pages

    WP.29-194-34e

    Uploaded by

    yosri baati

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    You are on page 1of 4

    Informal document WP.

    29-194-34
    194th WP.29, 12-15 November 2024
    Agenda item 2.3

    Artificial Intelligence in road


    vehicles
    Assessing AI
    • Artificial Intelligence (AI) is having an impact in the automotive industry on vehicle design,
    development and manufacturing.
     AI is enabling advanced capabilities on vehicles including safety-critical functions, such as perception,
    that is a crucial part of both advanced driver assistance systems (ADAS) and automated driving systems
    (ADS).

    • The use of AI challenges the application of established automotive safety practices (e.g. Functional
    Safety ISO 26262, Safety of the Intended Functionality ISO 21448). New techniques and standards are
    being developed to guide on safety assurance approaches when AI is used (e.g. ISO 8800).
     Several UNECE regulations require a functional safety assessment (UN-R13, UN-R13H, UN-R79, UN-
    R171, [UN-R(ADS)]…)

    • There is currently no UNECE framework for assessing AI when it is part of a safety critical system.
     Are existing provisions on e.g. functional safety in UN regulations appropriate for application to systems
    that incorporate AI in safety critical functions?
     What steps should we be taking to support Type Approval Authorities, and Technical Services, when they
    encounter AI and have to decide if its use is appropriate?
    UK work on safety assurance of Machine
    Learning
    In 2021, the UK’s Department for Transport commissioned a project1 which considered the safety assurance
    of vehicle systems including those with embedded models utilising machine learning.

    Based on this project, we undertook an exercise to develop a provisional set of requirements for systems that
    incorporate Machine Learning (ML) components. We identified several key considerations, including:
    • Identification of clear and robust requirements for the ML components (in the context of the target system)
    • Importance of data set management and data set specifications
    • Importance of appropriate selection of model architecture, training methods and final model selection
    • Importance of testing, evaluation, verification and validation
    • Importance of ongoing monitoring to confirm that the system continues to meet its performance targets

    1
    MIRA-AV-Safety-and-Security-Abridged-Report.pdf
    UK ask
    We ask that WP.29 establishes a dedicated informal working group on AI. Suggested activities:

    1. Build upon the existing activity that has considered the impact of AI and has delivered - “Considerations on
    Artificial Intelligence in the context of road vehicles” (WP.29-193-20).

    2. Support approval authorities and technical services in evaluating the use of AI, with a focus on safety-critical
    functions.

    3. Review the standards landscape for AI for road vehicles and how they interact with UN regulations.

    4. Make recommendations on further work for other (non-safety) vehicle systems.

    You might also like