Upgrade to Pro — share decks privately, control downloads, hide ads and more …

A4Q World Congress: 2021 IEEE Autonomous Driving AI Test Challenge

Exactpro
PRO
April 15, 2021
57

A4Q World Congress: 2021 IEEE Autonomous Driving AI Test Challenge

Elena Treshcheva - IEEE AV AI Test Challenge Organizing Committee,
Researcher, Exactpro

2021 IEEE Autonomous Driving AI Test Challenge
15.04.2021
--
To learn more about Exactpro, visit our website https://exactpro.com/

Follow us on
LinkedIn https://www.linkedin.com/company/exactpro-systems-llc
Twitter https://twitter.com/exactpro
Facebook https://www.facebook.com/exactpro/
Instagram https://www.instagram.com/exactpro/
Vkontakte https://vk.com/exactpro_llc

Subscribe to Exactpro YouTube channel https://www.youtube.com/c/exactprosystems

Exactpro
PRO

April 15, 2021
Tweet

Transcript

  1. 2021 IEEE Autonomous Driving
    AI Test Challenge
    Elena Treshcheva
    IEEE AV AI Test Challenge Organizing Committee,
    Researcher, Exactpro

    View Slide

  2. http://av-test-challenge.org/
    2021 IEEE Autonomous Driving AI Test Challenge

    View Slide

  3. - WHY both industry and academia need initiatives at the
    confluence of AI and software testing
    - WHAT the 2021 IEEE Autonomous Driving AI Test
    Challenge is all about
    - HOW one can explore the knowledge domain of testing
    complex intelligent systems
    Overview

    View Slide

  4. WHY: AI and Software Testing as a new field of studies
    Today’s context:
    - rapid advancements in AI-based technologies,
    active development of intelligent systems
    - research on trustworthiness, safety and reliability
    of AI
    - professional certifications and educational
    programs
    Indicators of the new area being established

    View Slide

  5. WHY: AI and Software Testing as a new field of studies

    View Slide

  6. AI systems as specific type of the system
    under test:
    - Non-deterministic
    - Non-testable with traditional methods
    AI to support software testing:
    - Oracle problem: telling right from wrong cannot be delegated to the algorithms
    - Limited applicability - not to testing per se, but rather to surrounding tasks
    A4Q AI and Software Testing Foundation Syllabus
    WHY: AI and Software Testing as a new field of studies

    View Slide

  7. WHY: AI and Software Testing as a new field of studies
    Quality characteristics of AI-based systems:
    - Ability to learn: the capacity of the system to learn from use for the
    system itself, or data and events it is exposed to
    - Ability to generalize: the ability of the system to apply to different
    and previously unseen scenarios
    - Trustworthiness: the degree to which the system is trusted by
    stakeholders
    A4Q AI and Software Testing Foundation Syllabus

    View Slide

  8. WHY: AI and Software Testing as a new field of studies
    Quality characteristics of AI-based systems:
    - Ability to learn: the capacity of the system to learn from use for the
    system itself, or data and events it is exposed to
    - Ability to generalize: the ability of the system to apply to different
    and previously unseen scenarios
    - Trustworthiness: the degree to which the system is trusted by
    stakeholders
    A4Q AI and Software Testing Foundation Syllabus

    View Slide

  9. WHY: AI and Software Testing as a new field of studies
    Quality characteristics of AI-based systems:
    - Ability to learn: the capacity of the system to learn from use for the
    system itself, or data and events it is exposed to
    - Ability to generalize: the ability of the system to apply to different
    and previously unseen scenarios
    - Trustworthiness: the degree to which the system is trusted by
    stakeholders
    A4Q AI and Software Testing Foundation Syllabus

    View Slide

  10. http://av-test-challenge.org/
    WHAT: IEEE Autonomous Driving AI Test Challenge

    View Slide

  11. http://ieeeaitests.com/
    The Third IEEE International Conference on Artificial
    Intelligence Testing

    View Slide

  12. http://www.ieeeaitests.com/html/callForPapers.html
    AI Test 2021 - Call for Papers:
    - Regular (8 pages) and Short papers (2
    pages)
    - AI Testing In Practice (8 pages)
    - Tool Demo Track (4 pages)
    The Third IEEE International Conference on Artificial
    Intelligence Testing

    View Slide

  13. http://www.ieeeaitests.com/html/callForPapers.html
    AI Test 2021 - Topics of Interest:
    - Testing AI applications
    - Applications of AI techniques to software testing
    - Data quality checking for AI applications
    The Third IEEE International Conference on Artificial
    Intelligence Testing

    View Slide

  14. http://av-test-challenge.org/
    WHAT: IEEE Autonomous Driving AI Test Challenge

    View Slide

  15. Autonomous Driving: Testing by simulation

    View Slide

  16. Autonomous Driving: Testing by simulation

    View Slide

  17. Autonomous Driving: Testing by simulation

    View Slide

  18. Autonomous Driving: Testing by simulation

    View Slide

  19. Autonomous Driving: Testing by simulation

    View Slide

  20. Autonomous Driving: Testing by simulation

    View Slide

  21. Autonomous Driving: Testing by simulation

    View Slide

  22. Autonomous Driving: Testing by simulation

    View Slide

  23. https://arxiv.org/pdf/2005.03778.pdf
    Rong et al. (2020) - LGSVL Simulator: A high fidelity
    simulator for autonomous driving

    View Slide

  24. https://arxiv.org/pdf/2005.03778.pdf
    Rong et al. (2020) - LGSVL Simulator: A high fidelity
    simulator for autonomous driving

    View Slide

  25. https://arxiv.org/pdf/2005.03778.pdf
    Rong et al. (2020) - LGSVL Simulator: A high fidelity
    simulator for autonomous driving

    View Slide

  26. Baidu Apollo Simulator

    View Slide

  27. Tutorial: End-to-end video tutorial of LGSVL simulator with Apollo driving
    Recordings and slides from the AV AI Test Challenge training sessions
    HOW: More on Software testing and AI

    View Slide

  28. Software testing for complex intelligent systems and autonomous vehicles -
    Playlist on the Exactpro YouTube channel
    Software Testing for Complex
    Intelligent Systems and
    Autonomous Vehicles
    Online Course
    HOW: More on Software testing and AI

    View Slide

  29. Online course from Exactpro
    Software Testing for Complex
    Intelligent Systems and
    Autonomous Vehicles
    Online Course
    - an introduction to AI-based software, risks associated with it
    and the necessity of its comprehensive testing
    - general knowledge of the software testing basics
    - an ability to use testing methods on real-life tasks
    - hands-on experience of using test tools and configuring a
    test environment
    - the skill of writing test cases, including automated test
    cases, and executing them
    - an ability to analyse and interpret test run results

    View Slide

  30. 1. Autonomous driving and software testing
    2. Introduction to self-driving cars
    3. Application programming interface
    4. Foundations of software testing
    5. Build Software to Test Software
    6. Reliability and safety
    7. Software defects and metrics
    8. Software negligence and testing coverage
    Online course from Exactpro
    Theoretical
    module
    Online course on the Exactpro
    YouTube channel

    View Slide

  31. 1. Launching the LGSVL Simulator
    2. Setting up a Python IDE
    3. Unit Testing
    4. Behavior Driven Development
    5. Data and Keyword Driven Testing
    6. Data Analysis and Reporting
    7. Running Apollo
    8. Modeling Test Scenarios
    Online course from Exactpro
    Practical
    module
    Online course on the Exactpro
    YouTube channel

    View Slide

  32. Takeaways
    - WHY: we need initiatives at the confluence of AI and software testing to
    work out new software testing techniques addressing sophisticated AI
    systems
    - WHAT: IEEE Autonomous Driving AI Test Challenge and AI Test
    Conference serve as a forum for discussion and development of new
    approaches to software testing
    - HOW: We definitely need to prepare software testers to work with
    intelligent systems. There are resources providing knowledge and
    training in the intersection of AI and Software Testing domains

    View Slide

  33. View Slide