Robustness is useless without reliable state information. For output feedback, a (\dot\hat\mathbfx = \mathbff(\hat\mathbfx,\mathbfu) + \mathbfL(\mathbfy - \hat\mathbfy)) with (\mathbfL) sufficiently large can exponentially recover estimated states. Sepulchre & Kokotović’s separation principle for nonlinear systems shows that a robust controller + high-gain observer preserves stability if the observer is fast enough.
When our mathematical "guess" of the system isn't 100% accurate. Robustness is useless without reliable state information
A technique that forces the system to "slide" along a predefined boundary of normal operation, making it incredibly resilient to disturbances. Input-to-State Stability (ISS): a (\dot\hat\mathbfx = \mathbff(\hat\mathbfx