Lecture 39 - Feedforward Control

Feedforward control is a strategy to reject persistent disturbances that cannot adequately be rejected with feedback control. So far, most of the focus of this course has been on feedback control. In certain situations, the performance of control systems can be enhanced greatly by the application of feedforward control. What you need to look for are two key characteristics:

1. An identifiable disturbance is affecting significantly the measured variable, in spite of the attempts of a feedback control system to regulate these effects
2. This disturbance can be measured, perhaps with the addition of instrumentation

• Lecture 39 Notes
• Lecture 39 Worksheet #1
• Lecture 39 Worksheet #2

Homework

1. Course reading for next class: 20 (PPC)
2. Assignment due by the start of Lecture #40: SP15

Relate each problem in the context of the overall course objectives.

Discussion Questions

Test the lecture's ideas before or after class - commit to an answer first, then check it with a classmate or an AI tutor.

1. Feedback control cannot act until the error exists; feedforward acts before it does. What two things must you have about the disturbance for feedforward to be possible at all, and what happens when one is missing?
2. The static feedforward gain is K_ff = -K_d/K_p. Derive the sign: why must the correction oppose the disturbance's effect, and what steady-state condition does this gain enforce exactly?
3. Feedforward is always installed alongside feedback, never instead of it. List three imperfections in practice that leave residual error for the feedback controller to clean up.
4. A "perfect" dynamic feedforward compensator sometimes requires predicting the future (more dead time on the manipulated path than the disturbance path). Explain why that compensator is physically unrealizable, and what the practical fallback is.

App: Compare pure feedback against feedback-plus-feedforward on the same disturbance in the Feedforward Cascade Control Studio - then deliberately corrupt the feedforward model (wrong gain) and watch feedback quietly absorb the residual.