Time | Title | Speaker |
---|---|---|
08:45 | Opening Remarks | |
09:00 | Invited Talk: Systematizing the Unusual: A Taxonomy-Driven Dataset of Edge Cases in Traffic | Krzysztof Czarnecki (University of Waterloo) |
09:30 | Invited Talk: Toward a Unified Framework for Runtime Monitoring and Self-Assessment in Autonomous Driving Systems | Thomas Griebel (Ulm University) |
10:00 | Invited Talk: The Smart-Data Approach to Testing and Validating | Arnaud de la Fortelle (heex.io) |
10:30 | Coffee Break | |
11:00 | Paper Presentation: Self-Supervised Pretraining for Aerial Road Extraction | Rupert Polley, Sai Vignesh Abishek Deenadayalan, J. Marius Zöllner |
11:30 | Paper Presentation: Balancing Progress and Safety: A Novel Risk-Aware Objective for RL in Autonomous Driving | Ahmed Abouelazm, Jonas Michel, Helen Gremmelmaier, Tim Joseph, Philip Schörner, J. Marius Zöllner |
12:00 | Panel Discussion: "Safety, Metrics, Benchmarks - Paths forward for AI Safety according to ISO/PAS 8800?" | |
12:30 | Lunch | |
13:30 | Follow up Workshop: Metrics4AI | https://mb4ad.ai/ |
One of the key challenges in assuring the safety of autonomous vehicles is their ability to operate reliably in the open world, where they may encounter rare and hazardous situations—commonly referred to as edge cases. Unlike routine driving, which involves frequent and repetitive behaviors that can be captured in relatively small datasets, edge cases are, by definition, infrequent and difficult to collect through limited-duration or small-scale data collection efforts.
EdgeScenes is a new dataset currently under development at the WISE Lab, designed to address this gap by systematically capturing and organizing edge case driving scenarios. It leverages a structured ontology to identify and categorize hazardous conditions related to road infrastructure, traffic behavior, foreign objects, and environmental factors. These scenarios are exemplified using crowdsourced video footage and annotated with a rich schema covering over 1,000 hazardous situations.
This talk will present an overview of the dataset's development, highlight key insights gained during its construction, and discuss the broad taxonomy of edge cases identified. While the scenarios originate from observations of human driving, they are directly relevant to the operational design domain of automated driving systems aiming to match or exceed human performance, particularly in safety-critical contexts. The dataset also supports the development and benchmarking of vision-language models that need to understand and reason about rare hazardous traffic conditions.
About:
Krzysztof Czarnecki is a Professor of Electrical and Computer Engineering and a University Research Chair at the University of Waterloo, where he leads the Waterloo Intelligent Systems Engineering (WISE) Laboratory. He currently also serves as an Associate Director of the Waterloo Centre for Automotive Research (WatCAR). His research focuses on assuring the safety of AI systems and driving behavior. In 2018, he co-led the development of the first autonomous vehicle tested on public roads in Canada. He has made significant contributions to automotive AI and software safety standards, including SAE J3164 and ISO 8800. Before joining the University of Waterloo, he worked at DaimlerChrysler Research in Germany (1995-2002), where he advanced software development practices and technologies for enterprise, automotive, and aerospace sectors. His work has been recognized with numerous awards, including the Premier's Research Excellence Award (2004) and the British Computing Society’s Upper Canada Award for Outstanding Contributions to the IT Industry (2008). He has also received twelve Best Paper Awards, two ACM Distinguished Paper Awards, and five Most Influential Paper Awards.
Self-assessment and runtime monitoring are essential for ensuring the safety and robustness of autonomous driving systems. This talk presents a unified framework for runtime monitoring and self-assessment, designed to observe and assess system performance in real time. The framework leverages subjective logic as a common interface to standardize the output of diverse self-assessment modules and enables their seamless combination. It is designed to be compatible with existing autonomous vehicle software stacks without requiring changes to functional components. By unifying and extending existing methods, this approach aims to enhance the overall safety, robustness, and transparency of autonomous driving systems.
About:
Thomas Griebel received the B.Sc. degree in mathematics and management from Ulm University, Germany, in 2015, the M.Sc. degree in applied mathematics from Missouri University of Science and Technology, USA, in 2017, and the M.Sc. degree in computational science and engineering from Ulm University, in 2019. Since 2019, he has been a researcher at the Institute of Measurement, Control and Microtechnology at Ulm University, and he completed his Dr. rer. nat. (Ph.D.) degree in 2025. His research interests include sensor data fusion, filtering and estimation, signal processing, and environment perception for autonomous driving. His special focus is on monitoring and self-assessment of tracking algorithms toward robust sensor fusion.
Heex has developed a Smart-Data platform, i.e. a solution enabling reconfigurable, event-based, data collection and processing. It provide a mechanism to target relevant events at the edge: triggers define events where valuable data has to be collected, e.g. hard brake, an unusual object or a successful avoidance maeuver. These events build specialized datasets that facilitate testing and validating within more specialized domains, with the aim to cover ODDS. Since this process is dynamic (sub-domains are validated and failures get confined to other subdomains), reconfigurability is supported. Moreover the solution provides good traceability for all these data, which may be critical (which software version produced events?). In summary, Smart-Data provides a powerful mechanism to build continuously evolving datasets from live systems.
About:
Arnaud de La Fortelle has engineering degrees from the French École Polytechnique and École des Ponts et Chaussées and a Ph.D. in Applied Mathematics. He managed (being coordinator twice) several French and European projects. Arnaud is currently CTO of heex.io, a French startup company focusing on smart-data solutions for autonomous systems. Before, he moved to MINES ParisTech in 2006 where he became director of the Center for Robotics in 2008. He was Visiting Professor at UC Berkeley in 2017-2018. He has been elected in 2009 to the Board of Governors of IEEE Intelligent Transportation System Society. He has been member of several program committees for conferences and was General Chair of IEEE Intelligent Vehicles Symposium 2019 in Paris. He was member, then president of the French ANR scientific evaluation committee for sustainable mobility and cities in 2008-2017. He also serves regularly as expert for the European research program (FP7, H2020). His main topic of interest is cooperative systems (perception, communication, data distribution, control, mathematical certification) and their applications (e.g. collective taxis, cooperative automated vehicles). He chairs the international research chair Drive for All with sponsors Valeo, Safran and Peugeot and partners UC Berkeley, EPFL and Shanghai Jiao Tong University.