Quantifying the safety of sensing, perception, reasoning, and guidance in autonomous systems