Speaker:Bowen Weng

Title: On the Possibilities and Impossibilities of Formal Robot Safety Testing

Abstract:

“A great problem of contemporary life is how to control the power of economic interests which ignore the harmful effects of their applied science and technology". This excerpt from Ralph Nader's groundbreaking book, "Unsafe at Any Speed: The Designed-in Dangers of American Automobiles," written in 1965, highlights a persistent conflict we face today with a variety of intelligent robots, autonomous systems, cyber-physical systems, and artificial intelligence (AI).

The resolution to this conflict depends significantly on the collective efforts of regulatory bodies, policy makers, standard organizations, community associations, and consumers. Their role is to justify, certify, falsify, verify, validate, and characterize the safety performance of the technology in an objective manner. This process is further strengthened by adopting a data-driven and experiment-oriented approach, namely testing. Despite the clear need for a formal safety testing approach that is unbiased, fair, nonhackable, interpretable, ethical, and efficient, achieving this goal has been, and to some extent, still remains a considerable challenge.

This talk is inspired to tackle this seemingly "mission impossible" in a variety of perspectives, covering the formal discussions of (i) safety performance metrics (how to define and measure safety?), (ii) scenario-based safety testing algorithms (how to provably test safety?), and (iii) adversarial safety testing (how to efficiently test safety?). These discussions present conditional formal safety testing alternatives demonstrated with intelligent vehicles and legged robots. Moreover, the talk will shed light on new challenges and impossibilities for the future, underscoring the ongoing need for innovative solutions in ensuring safety with advance robots, systems, and AI.

Bio: Dr. Bowen Weng is a research scientist at Transportation Research Center (TRC) Inc. on assignment to National Highway Traffic Safety Administration (NHTSA), where he leads the research and technical development of projects related to safety testing and performance evaluation of Automated Driving Systems (ADS). Dr. Weng received his M.S. in Electrical and Computer Engineering in 2016 from Case Western Reserve University, Cleveland, USA. He recently defended his Ph.D. dissertation at The Ohio State University, Columbus, USA. His research interests center around safety testing and safe autonomy, with applications to intelligent vehicles, legged robots, and multi-agent systems. Dr. Weng is the organizer of the IEEE-ITSC 2022, IEEE-IV 2023, and IEEE-ITSC 2023 workshops on “Safety Testing and Validation of Connected and Automated Vehicles” and the Associate Editor of IEEE-IV 2023. He also chairs the ASTM Committee F45.06 Legged Robot Systems for standard development.