Disturbance rejection

Disturbance rejection is the capacity of a control system to maintain its output at a desired setpoint despite the presence of external disturbances — forces, noises, or perturbations that would otherwise drive the system away from its target. It is one of the two primary objectives of control theory, alongside reference tracking, and it is the domain where the distinction between feedback and feedforward control matters most. Feedback rejects disturbances by detecting their effect on the output and counteracting; feedforward rejects them by predicting the disturbance and canceling it before it affects the output.

The measure of a controller's disturbance rejection capability is not how well it performs when there are no disturbances, but how small the output deviation remains when disturbances are large, sudden, and unmodeled. This is why disturbance rejection is the ultimate test of robustness: a controller that maintains performance across a wide range of disturbance types is a controller that has captured something true about the system's structure, not merely optimized for a particular scenario. The history of control failures — from pilot-induced oscillations in aircraft to the 2008 financial crisis — is largely a history of controllers with excellent nominal performance and poor disturbance rejection.