Acclimatizing the Operational Design Domain for Autonomous Driving Systems
Chen Sun, Zejian Deng, Wenbo Chu, Shen Li, Dongpu Cao
Abstract
The operational design domain (ODD) of an automated driving system (ADS) can be used to confine the environmental scope of where the ADS is safe to execute. ODD acclimatization is one of the necessary steps for validating vehicle safety in complex traffic environments. This article proposes an approach and architectural design to extract and enhance the ODD of the ADS based on the task scenario and the corresponding requirements in the development and verification cycle. The ODD is tightly focused on a unified quantifiable environmental model in the proposed approach while overseeing the ODD extraction process by formal specifications. In addition to the acclimatization framework, an implementation of the proposed approach is examined with two learning-based agents to demonstrate its feasibility. The proof of concept has shown promising directions for future work on ODD monitoring and on the applications in iterative development for ADSs.