Specifying Autonomous Driving Scenarios

Abstract

Defects in Autonomous driving systems (ADSs) might result in catastrophic losses of lives and properties. To avoid such defects, we need to first ensure high quality requirements, which highly possibly would lead to the delivery of high-quality ADSs. Specifying requirements for ADSs, a method needs to have terms/notations specific to ADSs such as complex traffic environments (e.g., pedestrians, roads). Use case modeling is commonly practiced in industry for requirements specification and modeling. In this paper, we propose a novel use case modeling methodology, named RUCM4ADS, which specializes the Restricted Use Case Modeling (RUCM). RUCM4ADS aims to specify ADS scenarios by integrating elements from both the autonomous driving domain and Operational World Model (OWM) Ontology. Accompanied with RUCM4ADS, we also develop an editor for it. To evaluate RUCM4ADS, we conducted one real-world case study with 10 use cases. We also conducted a preliminary controlled experiment, in a laboratory setting, to evaluate the applicability of RUCM4ADS. Results show that RUCM4ADS can be used for modeling ADS scenarios and has the potential to improve the overall applicability for specifying ADS scenarios as use case models.