You're responsible for validating a complex autonomous system that combines software, hardware, sensors, networking, and operator workflows. The system has to perform reliably in realistic conditions, but full end-to-end testing is expensive, slow, and sometimes hard to reproduce. You need a testing approach that gives confidence before release without overloading the team or missing high-risk failures.
What testing methodologies would you use for complex systems, and how would you decide the right mix of unit, integration, simulation, hardware-in-the-loop, end-to-end, and field testing?