
Validation of Safety Critical Systems: Introduction & Overview
Sydney Katz and Mykel Kochenderfer open Stanford's AA228V course on validating safety-critical systems, the engineering problem of proving that autonomous or automated systems behave safely before they are deployed. The lecture frames validation as an algorithmic challenge rather than a purely regulatory checklist, laying out why traditional testing breaks down for systems like self-driving cars and aircraft collision-avoidance software that face enormous input spaces. Katz and Kochenderfer survey the landscape of solution methods, including formal verification, falsification, and simulation-based testing, and preview how the course will build up algorithmic tools for estimating failure probabilities and finding rare but catastrophic failure modes. Drawing on their own research and the accompanying textbook, they set up the central tension the course will return to throughout: safety-critical systems must be trusted despite being too complex to exhaustively check. Sixty-seven minutes, delivered as a university course lecture with accompanying course materials.