Feedforward control

Feedback waits for an error to appear before reacting. Feedforward uses knowledge you already have, such as a model of the system, a planned motion profile, or a measured disturbance like outdoor temperature or supply voltage, to send the right command up front. Feedback then cleans up whatever the model got wrong.

Examples: a motion controller that adds the torque needed to accelerate a known mass along a planned trajectory; a heating controller that responds to a falling outdoor temperature before the room cools; a power stage that compensates for a measured change in input voltage.

Feedforward does not replace feedback, and it is only as good as the model behind it. A poor feedforward term makes the feedback loop work harder. A common practice is to get the feedback loop stable and robust first, then add feedforward to improve tracking.