Online Seminar

Artificial Intelligence in the Automotive Industry

13th & 14th September 2022

14:00 GMT Summer Time | 15:00 Europe Summer Time
09:00 Eastern Daylight Time | 06:00 Pacific Daylight Time

Artificial Intelligence (AI) and especially different aspects of Machine Learning (ML) are having an impact in many aspects of the Automotive Industry. The aim of this online seminar is to examine some of the uses of this technology and provide insight into how it can effectively be deployed.

Headline uses include the design process, inspection, testing and automated vehicle control. Although this seminar will focus on the Automotive environment and applications, many of the lessons will be transferrable to other industries. The aim of the seminar is to help delegates to appreciate:

Fundamental principles of ML
Variety of uses of AI / ML in the automotive and other industries
Opportunities for the use of ML especially for the complex tasks associated with Automated vehicles and driver assistance
A methodology for Assurance of Machine Learning Algorithms
How to approach the deployment of software based on ML in Automated Vehicles and the challenges from a regulatory perspective
Pitfalls and Good Practice in ML.

Agenda

Day 1 – Tuesday, 13^th September

14:00

Chairman's Introduction & Welcome
Ross Hughes, Vehicle Certification Agency

Unlocking AV2.0 with Simulation at Wayve
Vinh-Dieu Lam, Wayve

Introduction to NAFEMS Engineering Data Science Working Group
Fatma Kocer-Poyraz, Engineering Data Science Working Group Vice Chair

Update on ongoing projects and call for contributions
Engineering Data Science Working Group Member introductions
Member application presentations
Astrid Walle, Neural Concept Ltd.
Carsten Buchholz, Rolls Royce
Fatma Kocer-Poyraz, Altair
Peter Wooldridge, Monolith AI
Vladimir Balabanov, Boeing
Closing QA

16:00

Discussion Session

16:30

Close of Day 1

Day 2 – Wednesday, 14^th September

14:00

Chairman’s Overview of Day 1 and Introduction to Day 2
Ross Hughes, Vehicle Certification Agency

Machine Learning- Aided Anomaly Detection in Manufacturing
Mohammed Babakmehr, Ford

Assurance of Machine Learning in Autonomous Systems
Richard Hawkins, Assuring Autonomy International Programme

Enabling safe learning for AI-based planners in Automated Vehicles
Siddartha Khastgir, WMG - The University of Warwick

Expert Led Panel Discussion Session

16:20

Closing Remarks

16:50

Close of Day 2

Organised by

Presentation Abstracts

Assurance of machine learning in autonomous systems
Dr Richard Hawkins, Assuring Autonomy International Programme

Machine Learning (ML) is now used in a range of systems with results that are reported to exceed, under certain conditions, human performance. Many of these systems, in domains including automotive, will exhibit high degrees of autonomy and are safety critical. Establishing justified confidence in ML forms a core part of the safety case for these systems.

The Assuring Autonomy International Programme has developed a methodology for the Assurance of Machine Learning for use in Autonomous Systems (AMLAS). AMLAS comprises a process for systematically integrating safety assurance into the development of ML components and for generating the evidence base for explicitly justifying the acceptable safety of these components when integrated into autonomous system applications. AMLAS also provides a set of safety case patterns for linking the activities and artefacts of the process with an explicit safety case for the safety of the ML component.

In this presentation, we will introduce delegates to the AMLAS methodology and how it can be used to support the development of an explicit safety case for machine-learned components.

Machine Learning- Aided Anomaly Detection in Manufacturing
Mohammed Babakmehr, Ford

Standard tests are performed at the end of the production line on each unit/component manufactured in the automotive industry to ensure their quality before shipping to customers. The quality test procedures could be a highly complex experiment with many variables to monitor and consider. One of the popular data types vastly collected in these test procedures is multi-channel time-series data. Each time series represents a particular aspect or functionality of the product. In this talk, we will present the application of some of the state-of-the-art anomaly detection techniques for identifying defected units in automotive manufacturing lines over high-dimensional time series data. We will cover topics such as wavelet and Fourier analysis, statistical feature engineering, adaptive dimensionality reduction, unsupervised learning, deep learning, generative adversarial models, and their applications in quality time series analysis.