Project Summary
Control Barrier Functions (CBFs) provide formal safety guarantees for automated vehicles but often produce braking behaviors that feel unnatural to human drivers. This project aims to enhance CBFs to achieve more human-compatible braking while maintaining their rigorous safety properties. We will analyze phase-based car-following models to extract principles that characterize human-like deceleration patterns and apply these insights to redesign CBF parameters and tuning strategies. The research consists of three main components: (1) Systematic analysis of human-like braking principles from phase-based models and empirical driving data; (2) Development of enhanced CBF formulations with parameters specifically designed to replicate these principles; and (3) Comparative simulations to evaluate the enhanced CBF’s performance against standard approaches. The project directly addresses California’s automated vehicle deployment challenges by improving AV behavior at signalized intersections, reducing dilemma zone vulnerabilities, and creating more predictable interactions with human drivers. The results will deliver practical guidelines and methodologies for CBF-based controllers that better align with human expectations while preserving safety guarantees, ultimately accelerating the integration of connected and automated vehicles into existing traffic systems and improving transportation efficiency and safety throughout California.
