Project A04: Consistency of Hybrid/Continuous Models

This project advances the CRC’s goal of understanding, specifying, and operationalizing behavioral/functional consistency for hybrid CPS. It targets quantitative notions of consistency—measuring “model difference” in behaviors—to relate models, specifications, and the real system, enabling property transfer (e.g., by injecting quantified gaps as non-determinism in verification models). Because hybrid systems combine discrete and continuous dynamics with vast state spaces, the project integrates formal verification, testing (MiL/SiL/HiL), and runtime monitoring. A first focus domain is autonomous driving (KA‑RaceIng), covering views such as dynamics (ODEs), controllers, sensor/vehicle models, and computation delays. Metrics must reflect property-relevant variables (e.g., distance to track bounds, heading, velocity; minimal inter-vehicle distances). Scenario-based testing with heuristic search seeks behaviors minimizing safety margins to detect, quantify, and preserve functional consistency.