Glossary term
Feedforward Control
A control strategy that uses measured disturbances, commands, or models to act before an error is produced at the output.
Definition
methodA control strategy that uses measured disturbances, commands, or models to act before an error is produced at the output.
Feedforward control anticipates the required control action from known inputs or disturbances instead of waiting for feedback error. It is often combined with feedback, because feedforward can improve speed and disturbance rejection while feedback corrects model error and unmeasured effects.
Feedforward control computes a control action from a known command or measured disturbance before the controlled output deviates. If a disturbance can be measured early enough and its effect can be modelled, feedforward can cancel much of its impact before feedback has to react.
Engineering role
Feedforward is used in process plants, motion systems, robotics, temperature control, power converters, engine control, and precision positioning. Examples include adding torque proportional to desired acceleration in a servo axis, increasing fuel flow when load is known to rise, or adjusting a valve based on measured upstream flow disturbance.
Feedforward versus feedback
Feedback measures the result and corrects error. Feedforward predicts what input is needed. Feedforward can be fast because it does not wait for error, but it depends on model accuracy and disturbance measurement. Feedback is slower but more robust to unknown effects. In practice, high-quality control systems often use both.
Model requirements
Effective feedforward requires the correct sign, timing, gain, and dynamic shape. The compensation may include inverse plant dynamics, static gain maps, acceleration terms, gravity compensation, friction compensation, or measured disturbance paths. If dead time or actuator limits are present, the timing of the feedforward action becomes critical.
Implementation
Feedforward commands should be limited, filtered, and coordinated with actuator capability. When combined with PID control, anti-windup and bumpless transfer may be needed so that feedback does not fight the feedforward term. Commissioning should test both nominal operation and model mismatch.
Common mistakes
Common mistakes include treating feedforward as a replacement for feedback, using a model outside its valid operating range, and failing to account for measurement delay in the disturbance signal. Another error is applying an aggressive inverse model that demands actuator motion or power the real system cannot provide.