Conference Program




Day 1: Tuesday, October 22

Networking breakfast
8.15am - 8.50am

American Center for Mobility Introduction and Conference Welcome
8.50am - 9am

Komal Doshi
Director of mobility programs
Ann Arbor SPARK
USA

Keynote Session
9am - 12.30pm

Moderator

Komal Doshi
Director of mobility programs
Ann Arbor SPARK
USA

9am

Development of safety testing for automated driving systems

Michelle Chaka
Program director
Virginia Tech Transportation Institute
USA
Industry, government and other stakeholders are all working toward the common goal of saving lives and improving mobility through the use of safe, robust, reliable automated driving systems (ADS). Deployments of ADS are currently underway. However, safety testing for ADS is lacking for numerous reasons, including the facts that ubiquitous deployment is years away, the technology continues to evolve, the development is highly competitive and operational driving domains vary. The elements to realize safety testing, which could be widely accepted, are within reach. This presentation explores the various considerations involved in advancing the development of safety testing for ADS.

9.30am

Safety First for Automated Driving (SaFAD)

Neil Garbacik
ADAS system safety and security - systems and components – electrical engineering
FCA US LLC
USA
Through this collective work, we define 12 guiding principles of safe automated driving development for SAE Level 3 and Level 4 automated driving systems, derived from a comprehensive collection of publications and recommendations from public authorities and consumer associations. The work expands to apply the principles to the development lifecycle of automated driving technology, covering requirements development through verification and validation and continuing beyond product deployment. The guidance provided in the paper defines a base framework to initiate industry alignment on developing a safe automated driving vehicle, benefiting everyone from startups to OEMs.

10am

Automated vehicles working together with CARMA3

Dr Taylor Lochrane
Technical program manager
USDOT Federal Highway Administration
USA
Cooperative Driving Automation (CDA) supports and enables automated vehicles to cooperate through communication between vehicles, infrastructure devices capable of communication, and road users such as pedestrians, bicyclists and scooter riders. The Federal Highway Administration (FHWA) developed the CARMA Platform and CARMA Cloud to support the research and development of CDA features in support of transportation systems management and operations (TSMO). The CARMA Platform enables cooperative research functionality to an automated driving system (ADS), and CARMA Cloud enables the roadway to provide information to support and enable safe operation for new TSMO strategies. Developed to be vehicle and technology agnostic, CARMA was designed using open-source software (OSS) and an agile software development process. The latest version of CARMA is available on GitHub and open for collaboration. As FHWA highlights the latest version of CARMA (CARMA3) and collaboration opportunities, learn how CARMA will transform transportation, improving efficiency and safety through automated vehicles working together.

10.30am - 11am

Break

11am

Toward safe and socially acceptable autonomous vehicles

Dr Liam Pedersen
Deputy director - robotics
Alliance Innovation Lab - Silicon Valley
USA
In the future, autonomous vehicles must not only aim to be safe, but also behave in a socially appropriate manner to be accepted in our society. In fact, safety and social acceptability are tightly linked, as inappropriate or unpredictable behaviors on the part of a car will engender potentially unsafe actions by other road users. This presentation will outline the use of collective human and machine intelligences to solve these difficult problems.

11.30am

Secure, trusted and scalable platform to accelerate AV test and development

Mitra Sinha
Autonomous development lead
Microsoft Azure
USA
Autonomous driving is fundamentally transforming the transportation industry, with computer vision, AI and HPC leading the change. The data streams generated by autonomous vehicles are unprecedented, resulting in the need for massive scale across the entire workflow: from PB-scale data ingest and storage to end-to-end algorithm validation, simulation and training. In this session we will show how a hyper-scale public cloud like Microsoft Azure can provide a secure, trusted and scalable platform to help auto makers scale their validation and training jobs and gain development process efficiencies for faster time-to-market.

12pm

Verification and validation to ensure safety first for automated driving

Dr Oliver Rumpf-Steppat
Head of department, product requirements, development and connected drive
BMW of North America
USA
This presentation addresses the verification and validation (V&V) of automated driving systems, including field monitoring and updates. It introduces the main steps and general approach, and defines the scope. It continues with an overview of the five key challenges that make V&V unique to L3 and higher automated driving systems. Solutions are proposed for each of the challenges and include a discussion of the various test platforms involved. The presentation also includes a discussion about the quantity and quality of real-world driving required and the use of simulation for V&V. Finally, the presentation focuses on specific V&V considerations for individual elements of an automated driving system. Although this presentation recognizes the possibility that validation testing may trigger functional design changes, most of it focuses on validating a stable system in a fixed ODD. However, it discusses post-deployment field operations, including the monitoring and management of configuration and ODD changes and updates.

12.30pm - 2pm

Lunch

Test, Development and Validation – Innovations and Best Practices to Ensure Safety
2pm - 5pm

Moderator

Komal Doshi
Director of mobility programs
Ann Arbor SPARK
USA

2pm

An open, transparent, industry-driven approach to AV safety

Jack Weast
Senior principal engineer, Intel & VP autonomous vehicle standards, Mobileye
Intel Corporation
USA
At Intel and Mobileye, saving lives drives us. Since joining forces, we’ve spread the word on the need for a safety standard for autonomous vehicles (AV), and how consumers and regulators alike demand transparency not offered by existing metrics used in AV safety claims. We proposed Responsibility-Sensitive Safety as a potential solution, a formal, mathematical model that defines what safe driving looks like. It was our first step toward building consensus in the industry. Today we take the next step in that journey, diving deeper into the makeup of RSS: what is this model, how does it work under the hood, and how can RSS help us balance the trade-off between the safety and usefulness of AVs? Higher levels of safety may result in overly conservative AVs that nobody wants on the road. So where should industry and the public draw the line to answer the question, “How safe is safe enough”? Help us drive the conversation today that will enable the autonomous tomorrow.

2.30pm

Realizing measurable safety using scenarios and coverage-driven verification

Yoav Hollander
Founder and CTO
Foretellix Ltd
ISRAEL
In this presentation, Yoav Hollander describes an approach to address the largest barrier to the broad deployment of autonomous vehicles: measurable safety. It is now common practice to use scenarios as one of the main ways to ensure AV safety. The presentation will emphasize the need for scenarios to be portable across the various execution platforms used (test tracks, specific simulators, etc.), the various stakeholders (OEMs, subsystem creators, regulators, etc.), various use modes (fully random and deterministic) and so on. It will also discuss how coverage and grading metrics play a crucial role in defining scenarios, evaluating the results and ensuring safety. Coverage-driven verification offers a revolutionary approach, shifting the industry from looking at the number of miles driven to test AVs, to using a new measure – quality of coverage. With this new measure, the industry can generate well-understood, comparable, communicable metrics. These metrics are the first steps toward quantifying the safety of autonomous vehicles.

3pm

Practical validation of AI within the SOTIF framework

Dr Edward Schwalb
Lead scientist
MSC Software
USA
Whereas control on individual outcomes for autonomous vehicles is limited, engineering processes must exert control over the continuous improvement of overall performance statistics. The ISO/PAS 21448 SOTIF standard partitions scenarios into safe vs unsafe, known vs unknown, and recommends continuous improvement to maximize the portion of known safe scenarios at the expense of the other partitions. The required quantitative analysis is performed by combining residual risk analysis, hazard modeling and Bayesian probabilistic reasoning to encompass multiple agents. We detail methods for modeling hazards and leverage those models within a driving loop in which perception algorithms are trained to detect hazards, and the decision logic actively avoids accidents. Finally, we show how to perform engineering against unknown scenarios quantitatively, and to discriminate incremental improvements from regression.

3.30pm - 4pm

Break

4pm - 5pm

Panel Discussion: Setting the standards for safety testing and development

How can the industry best take safety to the next level? Is a collaborative approach to testing and development the best solution?
Neil Garbacik
ADAS system safety and security - systems and components – electrical engineering
FCA US LLC
USA
Yoav Hollander
Founder and CTO
Foretellix Ltd
ISRAEL
David Woessner
Executive vice president of corporate development and regulatory affairs
LM Industries
USA
Dr Taylor Lochrane
Technical program manager
USDOT Federal Highway Administration
USA
Moderator:
Komal Doshi , director of mobility programs, Ann Arbor SPARK

Day 2: Wednesday, October 23

Solving Complex Challenges – Simulation, High-Performance Computing, Testing Software and AI
9am - 5pm

Moderator

Dr Edward Schwalb
Lead scientist
MSC Software
USA

9am

A massive simulation approach to verify and validate AV systems

Tony Gioutsos
Director portfolio development for autonomous Americas
Siemens
USA
This presentation discusses a massive validation and verification framework for ADAS and autonomous vehicles. The framework combines data management, test automation and results post-processing capabilities into a seamless workflow. For the execution, highly accurate simulation solutions are coupled to model the vehicle, the sensors and the environment, with the opportunity to exchange component models with different fidelity levels, depending on the exact test case. By coupling results to requirements and back, the approach allows for detailed analysis of AV systems. Cloud or cluster processing is also discussed, as well as coupling to real-world test tracks.

9.30am

Challenges of deep learning in the automotive industry

John Manera
Field automotive CTO
Dell EMC
USA
Larry Vivolo
Senior business development manager - automotive and EDA
Dell EMC
USA
After briefly introducing deep learning, the talk will focus on the common workflow of constructing a neural network in terms of the lifecycle of automotive product development: the data collection and acquisition phase, data annotation phase, quality checks and finally constructing the network with its test and validation as a last step. The audience will understand the common workflow and the basics of constructing a deep-learning-based classifier for automotive product development; become aware of typical challenges/problems and how to avoid and counter them; learn how to test and validate deep-learning-based algorithms for autonomous driving; and understand how many miles must be driven, how many images annotated, and the massive investment needed in terms of effort for test and validation.

10am

Bringing accuracy to autonomous vehicles simulation with high-fidelity physics

Dr Sandeep Sovani
Global director, automotive industry
Ansys Inc
USA
This talk will provide insights into high-fidelity physics-based simulation methods used in autonomous vehicles to drive scenario simulation as well as detailed component development. In particular, light simulation for optical sensors (camera, lidar) and electromagnetics simulation for radar will be discussed. Additionally, reduced order models (ROMs) and modeling methods will be presented. ROMs are used for expediting simulation beyond real time, while maintaining a high degree of physical accuracy. Finally, driving scenario simulation with physically accurate sensor models will be presented. Examples of typical corner cases that can only be addressed with high-fidelity physics-based simulation will be shown.

10.30am - 11am

Break

11am

Automating simulation for safer self-driving

Ferenc Pintér
AiSim product manager
AImotive
HUNGARY
The presentation details how automated simulation testing can accelerate the development of automated driving solutions, while making them safer. The importance of simulation is showcased by examining points of interest in AImotive’s development pipeline. First, while detailing the demands of simulation for autonomous driving, the need for a comprehensive content library is stressed. Second, the uses and limitations of simulated training data for neural networks are touched upon, followed by how scenarios should be defined based on real-world situations, alongside functional safety engineering. Finally, two case studies provide insight into how simulation has solved problems in AImotive’s internal development efforts.

11.30am

Scenario-based virtual validation of AVs

Dr Henning Lategahn
CEO
Atlatec GmbH
GERMANY
Waymo, Cruise and Zoox lead the pack in autonomous driving by some standards. The one thing they have in common is the heavy use of simulations in which real-world data is processed, and extracted scenario descriptions are fed into a simulator. The simulator replays these scenarios time and time again, leading to a quantum leap in training the underlying AI. We present a way to extract a high-fidelity road model and traffic scenarios from real-world data. This scenario description can thereafter be used in ADAS simulators such as CarMaker, PreScan, Virtual Test Drive and more, and we thereby contribute to a vast scenario-based virtual validation strategy.

12pm

The AV test fleet of the future is virtual

Norm Marks
Global senior director, automotive – enterprise
Nvidia
USA
Simulation has proved to be essential for safely testing and validating self-driving technology before it’s deployed on the road. This presentation will delve into existing simulation challenges and why the industry needs a better solution to test and validate AVs than what has been previously available. Learn how to build a unique virtual AV test fleet in the cloud, and utilize an open, scalable, bit accurate, hardware-in-the-loop solution that will allow the development and validation of AVs without putting others on the road in harm’s way. This ultimately enables greater efficiency, cost-effectiveness and safety than what is physically possible to achieve with real-world test drives.

12.30pm - 2pm

Lunch

2pm

Scenario-driven development 2.0

Philipp Renner
Technical sales manager
Understand.ai
GERMANY
Given its complexity, the testing of Level 3 and 4 highly automated driving has become the bottleneck. The old paradigm of mileage-driven testing will need to be replaced by scenario-based testing to test autonomous vehicles in a measurable, consistent and deterministic way. The new paradigm – scenario-driven development – is most efficient when adopted early in the process of design and development of HAD. In this presentation we define ​scenarios and explain why they are important. We illustrate at what points scenarios create value in the value chain of autonomous vehicle design and deployment. We also present our pipeline for scenarios.

2.30pm

Virtual validation and simulation at large scale, for training, testing and deploying automated driving systems

Heikki Laine
VP product and marketing
Cognata Ltd
USA
Machine learning and deep neural networks require tremendous quantities of data for training and validation, but even at scale, raw, repetitive or inaccurately labeled data doesn’t produce results. Training and validation call for accurate, large-scale data sets comprising common scenarios, edge cases and every sort of variation in between. Furthermore, each process requires a distinct data set. We will explore how new techniques in synthetic data generation are helping time-pressured industries like automated driving satisfy the ever-growing need for larger, more diverse and highly accurate data sets.

3pm

Autonomous vehicles: automation of training, simulation and perception detection using high-performance computing

Srijani Dey
Chief architect, Americas Analytics Center of Excellence lead
DXC Technology
USA
This session will explore hyperscaling with data and a compute platform to sustain the increasing data ingestion with SLAs and KPIs and enable data access in different formats with increasing performance for scenario detection, tagging and labeling. Additionally, we will look at the automation capability of the platform and the handling of containerized workloads.

3.30pm - 4pm

Break

4pm - 5pm

Panel Discussion: Bridging the gap – simulation frameworks and real-world validation methodologies

Simulation is a pivotal enabler for the development of automated vehicles within reasonable timelines. This panel will provide an overview of current methodologies and recent advancements in simulation, as well as approaches to validate simulated results with real-world results. Toolchains and pipelines for scene and scenario generation (both creation and extraction) will also be presented. Multi-fidelity simulation (full sensor vs. object level) and its relationship to scene design will be discussed. Standards and regulation programs and adoption rates will also be touched on throughout the panel.
Steve Rotenberg
CEO
VectorZero
USA
Jace Allen
Business development manager – simulation, test and EEDM
dSPACE Inc
USA
Natalie Afonina
Product lead HD maps
Mapbox
USA
Moderator:
Daniel Schambach, co-founder/head of design, Metamoto Inc

Day 3: Thursday, October 24

Validation in the Virtual Domain
9am - 10.30am

Moderator

Vivek Jaikamal
Business development manager
AVL Test Systems Inc
USA

9am

Validation of autonomous safety in the context of SOTIF

Jace Allen
Business development manager – simulation, test and EEDM
dSPACE Inc
USA
For the validation of functional safety for automotive E/E systems and embedded software, proper workflows and virtual test methods are necessary. The need for virtual testing and efficient collaboration is increasing because of new challenges with the development of advanced driver assistance systems (ADAS) and autonomous driving (AD). Also, the requirements for a reliable virtual testing process are increasing as ADAS/AD systems are becoming more safety critical. The presentation outlines necessary V&V strategies that are compliant with ISO 26262 and ISO/PAS 21448 to ensure the proper functionality and safety goals of ADAS/AD systems.

9.30am

Automating autonomous vehicle requirements development and validation

Rick Sturgeon
Senior director, transportation and mobility
Dassault Systèmes
USA
The validation process for autonomous vehicle embedded systems faces a serious challenge: the vehicle makes decisions in very unpredictable environments, which means the level of testing must be very high. As test objectives are derived from the functional requirements written using natural language, each and every functional test must be created manually, which greatly reduces the level of confidence necessary for an autonomous vehicle. With Dassault Systèmes’ STIMULUS, users are able to simulate, debug and validate the requirement first and then automatically test the actual system against its functional specification, using a large number of tests generated automatically.

10am

Making machine perception real with high-fidelity synthetic data

Victor Gonzalez
CEO
Anyverse (Next Limit SL)
SPAIN
Training self-driving technology is a crucial step in autonomous vehicle development and deployment, especially in terms of the much-scrutinized safety issue. Unfortunately, this part of the process is still facing a major data challenge. The real-world approach has proved to be insufficient and time-consuming, slowing down the progress and exposing it to numerous loopholes. An alternative solution is the use of virtual images that meet specific training and testing needs and complement real-world data. However, synthetic data is not made equal. It needs to be as real and physically accurate as possible, and include all segmentation data. This is Anyverse.

10.30am - 11am

Break

Data and Connectivity. Testing AV Performance
11am - 12.30pm

Moderator

Vivek Jaikamal
Business development manager
AVL Test Systems Inc
USA

11am

Chassis dynamometer testing methodology development for CAV energy consumption

Kevin Stutenberg
Principal research engineer
Argonne National Laboratory
USA
As connected and automated vehicle (CAV) technologies begin to contribute to a greater share of vehicle miles traveled, methods for evaluation of these systems must advance to accurately quantify their impact. Chassis dynamometers have long been a standard for vehicle testing and development activities, in particular for the evaluation of energy use, fuel consumption and criteria emissions, providing an environment in which a vehicle may be tested in safe, variable, controllable and repeatable conditions with flexible instrumentation. The focus of this work will be on the development of methods for the characterization of CAV behavior and impact on vehicle performance.

11.30am

Motion planning at the physical limits

Dr Stefano Longo
Head of automotive
Embotech
SWITZERLAND
A vehicle’s physical capabilities are crucial for the feasibility and smoothness of any maneuver. Traditional motion planning methods for AD neglect most of the physics, being conservative or requiring advanced low-level vehicle controls that are often not present or are prohibitively expensive. We demonstrate physics-based motion planning technology, using numerical optimization, to calculate smooth and safe trajectories that can be easily followed by standard low-level vehicle controllers. Based on recent advances in embedded optimization technology, we capture most of the relevant vehicle dynamics while driving on a highway or on rural roads, significantly extending the performance envelope of autonomous cars.

12pm

"Can't we all just get along?" An approach to ensuring interoperability in V2X hardware

Kimberly Clavin
VP of engineering
Loop by Pillar (part of Accenture Industry X.0 company)
USA
Nick Hegemier
Managing director of infrastructure
DriveOhio
USA
Using traditional agile software practices such as test-driven development and continuous integration within a closed-loop ecosystem of V2X hardware, one can validate, in high quantities, the authenticity and accuracy of messages being tossed between the city and cars as well as between the cars themselves. This talk will demonstrate a case study of work performed at DriveOhio. DriveOhio, together with Pillar, a part of Accenture's Industry X.0, has created a path by which to test the interoperability of hardware equipment within a lab environment. This system allows for application-based test scenarios in an automated fashion. These tests can be run in high-quantity batches, which is a great advantage over driving roads for single data points. In essence, this ecosystem allows for changes in standards and/or technology to be easily tested, and errors rectified quickly.

12.30pm - 2pm

Lunch

2pm - 3pm

Panel Discussion: Expediting the AV future – the next phase of the Silicon Valley-Detroit relationship

Engineers, analysts, commentators and automotive experts in both Detroit and Silicon Valley agree that the two regions must work together to speed up the development and deployment of autonomous vehicles (AVs). This panel will explore how OEMs and Silicon Valley software companies have been collaborating on the most pressing technical and production-oriented challenges. What are the biggest obstacles facing both sets of companies? When should Silicon Valley companies consider opening Detroit offices and vice versa? What does this collaboration mean for the future of the American automotive industry?
Aditya Srinivasan
General manager, North America
Innoviz Technologies
USA
Steve Vozar
CTO
May Mobility
USA
Mohammad Musa
Co-founder and CEO
Deepen AI
USA
Moderator:
Qasar Younis, co-founder and CEO, Applied Intuition Inc
Please note: this conference program may be subject to change.

AUTONOMOUS VEHICLE INTERNATIONAL

Autonomous Vehicle International magazine

SUBSCRIBE TO THE MAGAZINE FOR FREE!