The Robust Maximum Principle