From High-Dimensional Spaces to Verifiable ODD Coverage for Safety-Critical AI-based Systems

Thomas Stefani, Johann Maximilian Christensen, Elena Hoemann, Frank Köster, Sven Hallerbach|April 2, 2026arXiv

Key Takeaway

To certify AI in safety-critical domains like aviation, you need a structured way to prove complete test coverage across all operating conditions; this paper provides an engineering method to do that at scale using parameter discretization and criticality-based filtering.

Summary

This paper tackles a critical certification challenge: proving that AI systems used in aviation have been tested across all relevant operating conditions.

safety evaluation applications

Key Terms

operational-design-domain safety-critical-systems coverage-verification dimension-reduction