A reproducible, data-driven framework for safety and functionality verification of Level-4 automated buses through real-world testing

Abstract

Ensuring the safety of automated buses is a prerequisite for their deployment in real-world public transportation. This study focuses on the systematic verification and validation of an autonomous bus using established Advanced Driver Assistance System (ADAS) functions, rather than introducing new functionalities. We propose a Functionality and Safety Verification, Analysis, and Assessment (FSVAA) framework that establishes a structured transition from the Operational Design Domain (ODD) to Dynamic Driving Task (DDT) elements and corresponding test scenarios. The framework is implemented on a full-size Level-4 automated bus equipped with ADASTEC’s flowride.ai platform. A total of 340 on-track tests were conducted across Automated Emergency Braking (AEB), Automated Trajectory Following (ATF), and Adaptive Cruise Control (ACC). System performance was evaluated using safety- and comfort-related indicators, including time-to-collision (TTC), time headway (THW), acceleration, and jerk. The results demonstrate consistent pass/fail decisions and support compliance with acceptance criteria required for progression toward open-road testing, providing a reproducible foundation for continuous safety validation of automated buses.