Unknown-Unsafe Scenario Generation for Verification and Validation of Automated Vehicles

The verification and validation of automated vehicles is a mammoth task, and is gaining increasing importance with the increasing number of automated vehicles coming into production. Scenario-based assessment is among the most popular approaches to validate the safety of automated vehicles. This work proposes a scenario generation approach in line with the ISO 21448 standard, Safety Of The Intended Functionality (SOTIF). A novelty metric is proposed, and is used along with a criticality metric in an optimization framework to systematically explore the unknown unsafe scenarios as per SOTIF. The proposed approach is applied to a complex scenario example with numerous interacting actors, and the resulting novel critical scenarios are categorised and analyzed. The developed framework can be extended and applied for the SOTIF validation of highly automated vehicles.