The goal of ensuring safe mobility has been at the forefront of government agencies and industry since the introduction of the mechanized transportation. Connected vehicle technologies were initially introduced to allow vehicle to vehicle and vehicle to infrastructure systems to enable the next generation of safe vehicles. Similarly, autonomous vehicles has often been looked at as a way to reduce accidents and fatalities, often associated with human error. However, fulfilling the promise of vision zero (zero deaths, accidents, etc.) is still a long way off (or maybe not possible at all?). This workshop is aimed at understanding if it is possible to build provably safe vehicles that operate in the presence of humans, both as passengers of the AVs and as pedestrians, bicyclists, etc. Techniques from the formal methods communities have already been successfully applied to a number of cyber-physical systems including industrial automation and the aerospace industry, however autonomous driving presents a new generation of problems. Specifically, formal methods have been quite successful in the past at analyzing systems (e.g., controls systems) which do not learn over time and do not directly interface with complex human interactions. In contrast to today’s promising advances in machine learning based sensing and driving best exemplified by many of the recent impressive technology demonstrations, ensuring safety for learning-based systems is an open question. This workshop is aimed at bringing together the formal methods community and the transportation community to understand how to build safe systems in the extremely complex driving environments where humans and robots will interact.
This workshop will include a poster session; a request for posters will be sent to registered participants in advance of the workshop.
(University of California, Berkeley (UC Berkeley), CITRIS)
Lillian Ratliff (University of Washington)
Richard Sowers (University of Illinois at Urbana-Champaign)
Jonathan Sprinkle (University of Arizona)
Daniel Work (Vanderbilt University)