Concept and Metamodel to Support Cross-Domain Safety Analysis for ODD Expansion of Autonomous Systems.

Automated driving systems (ADS) can improve efficiency in logistics and last-mile delivery, but a major challenge is ensuring safety for operational design domain (ODD) expansion or cross-domain deployment. Various ontologies and formats exist for modeling and representing the operational environment. However, their structuring schemes are not suitable for safety engineering activities, as the safety-relevant aspects of the environment differ from those relevant for other purposes, e.g., simulation scenario representation. This paper addresses the problem of effectively supporting safety engineers in performing environmental safety analyses considering cross-domain aspects and the impact of environmental changes. We contribute a concept for modeling and comparing operational design domains as well as algorithms for semi-automatically analyzing change impact. The approach is model-based, integrated within the Digital Dependability Identity (DDI) framework, and has been evaluated qualitatively for ADS cross-deployment from a logistic warehouse to urban environments. The evaluation suggests that the approach is a suitable starting point for explicitly linking ontological domain modeling with safety engineering. It also helps safety engineers to think about ODDs in a structured way, performing change impact analyses regarding specification gaps, and enabling cross-domain learning.