From Breaking to Fixing: Enforcing and Repairing Autonomous Vehicle Compliance with Traffic Laws
Abstract:
Autonomous vehicles (AVs) must not only avoid accidents but also adhere to traffic laws, which can vary significantly across countries. In this talk, I present a line of work addressing this challenge through specification, enforcement, and repair. Starting with LawBreaker, we develop a domain-specific language and fuzzing approach for exposing ADS violations of real-world traffic laws. Next, REDriver introduces a runtime enforcement framework that monitors ADS-planned trajectories and adjusts them when violations are predicted, preserving safety with minimal disruption. Finally, FixDrive leverages multimodal large language models (MLLMs) to generate high-level, interpretable 'repairs' to driving strategies based on past violations, enabling AVs to adapt offline with minimal cost. Together, these approaches offer a principled and practical path towards safer and more law-compliant autonomous vehicles.
Bio:
Chris Poskitt (https://cposkitt.github.io/) is a faculty member at Singapore Management University, where is part of the Centre for Research on Intelligent Software Engineering. Prior to Singapore, he undertook his doctoral studies at the University of York and held a postdoctoral research position at ETH Zürich. His research broadly addresses the problem of engineering correct and secure software, especially in the context of cyber-physical systems (e.g. industrial control systems, autonomous vehicles). In addition to software engineering, his interests span formal methods, cybersecurity, and computer science education.