Virtual Autonomous Driving Meetup #3

Event Password: ADmeetupV3

We are continuing our 3rd virtual Meetup on Autonomous Driving on July 30th!

Our special take on the matter will certainly remain: the Meetup aims to provide a broad overview, with talks from software companies, startups, OEMs and Tiers; addressing anything from convolutional neural networks to safety and legal questions.

The link to the web conference will be made accessible to everyone who made it to the attendees list. 😉

___ Agenda ___

_ Introduction & Welcome

_Session 1: Autonomous Driving and Open Source - Is this a good idea?
Andy Riexinger, Product Manager Autonomous Driving, Robert Bosch GmbH
www.bosch.de

• Challenges of a very complex AD software development
• Usage of Open Source technolgies and their challenges
• Open source software and the automotive industry
• Collaboration and current projects

_ Session 2: Edge Case Research: Towards Responsible Deployment of Autonomous Vehicles
Mark Shepherd, Edge Case Research
www.edge-case-research.com

The critical decision about when autonomous vehicles are ready to deploy is likely to be made with insufficient amounts of data to assure a safety outcome. However, it is possible to make a responsible deployment decision despite this uncertainty. It involves a combination of engineering rigor, validation, and post-deployment feedback all founded on a strong safety culture. An example of this comes from our work on automated testing of perception systems that helps finding edge cases and potentially unsafe behavior of AD perception software both before and after deployment.

_ Session 3: Enabling Verification and Certification of Autonomous driving function – A Formal Verification Approach
Andreas Lauringer, Jinwei Zhou, Kontrol GmbH
www.kontrol.tech

Ensuring safety is one of the key aspects for the introduction of autonomous driving. Empirical methods rely on large amounts of data from measurements or simulations, which create test environments for SUTs. Unfortunately, it is impossible to measure or simulate all possible situations, thus limiting the applicability of such methods. Therefore, we propose an alternative approach that monitors a vehicle’s behavior and verifies whether it obeys all rules, regulations and requirements. This approach is based on formal verification, which, in turn, enables certification of automation systems. Our methodology starts with a translation of traffic rules, which point out safe, and legal behaviors. We use this information as well as predicted future states of a vehicle over a predefined time horizon to verify whether the vehicle behavior violates any rules in this time horizon. Our solution enables the verification of the SUT and certifies the decision and the planned trajectory of a vehicle

_ Session 4: Towards Safe Autonomous Driving: Capture Uncertainty in Object Detection using Deep Learning Approach
Di Feng, Doctoral researcher at Bosch Research and Ulm University
www.bosch.com/de/forschung/

• An overview of perceptual uncertainties in autonomous driving
• Uncertainty modelling and applications in LiDAR-based 3D object detection

___ Further Details ___

_ The talks are all held in English and will take ~15 minutes each + ~10 minutes Q&A
_ Sponsor: Warmest thanks to Mathworks for enabling this online Meetup event by providing the web conferencing tool!
_ Exact time slots for the presentations may be adapted