Knowledge-Based Derivation of Relevant Scenario Objects for Testing Automated On-Sight Train Operations

The testing of automated rail vehicles presents substantial challenges. This is particularly evident in the context of automated on sight train operation, where the reliability of perception systems must be assessed through rigorous and systematic testing. Scenario based testing, which is frequently applied for this purpose, requires fundamental information regarding relevant scenario parameters and potential events. In the railway domain, the availability of extensive real world operational data is limited, which makes the structured use of knowledge essential. This work uses the example of an automated hump locomotive, representing an automated on sight driving railway vehicle, to demonstrate how relevant knowledge entities and derived scenario objects can be extracted from a corpus of knowledge sources while considering the system specific operational design domain. Suitable sources are defined, relevant knowledge entities are identified, and a dataset of scenario objects is established. The evaluation of the object list is carried out using an analytical approach based on object occurrence frequency. The resulting dataset of scenario objects can be integrated into knowledge based scenario generation and supports the targeted derivation of scenarios tailored to the specific operational design domain.