Virtual Continuous Testbeds with LLM-Based Test Case Generation: Toward Closed-Loop Validation of Safety-Critical Systems
: This paper presents an automated approach to deriving informal test cases from Requirements Interchange Format (ReqIF)-structured system requirements based on general-purpose Large Language Models (LLMs) with integrated linkage to test evidences based on Digital Dependability Identities (DDIs). Our method enables scenario-based testing while drastically reducing effort and cost: test cases become available within hours instead of months. The automation supports formal traceability, safety argumentation, and change impact analysis. An automated conversion of derived test cases into ASAM OpenScenario (XOSC) and Open Test Sequence Exchange (OTX) formats accelerates design and implementation of platform-independent, executable test specifications while improving their reuse across development stages. A systematic review process allows domain experts to refine and extend generated specifications. We discuss testbed configuration including resource allocation, toolchain integration, and parameter initialization to satisfy test constraints and strengthen confidence in results against the original acceptance criteria. Finally, we execute the reviewed test cases in a simulated environment within a Virtual Continuous Testing (VCT) pipeline for efficient verification and validation in an X-in-the-Loop (XiL) testbed based on VCIP/FERAL. The closed-loop approach advances automated testing by combining efficiency, consistency, and scalability in test generation, which is showcased within an automotive use case.