Feedforward and Feedback Control

Feedforward applies the planned command in advance; feedback corrects the difference between plan and reality. Good control uses both.

Two complementary strategies

Feedforward control applies a pre-computed command based on the plan, without waiting to see the result. Feedback control measures the result and corrects the error. Feedforward is fast and anticipatory but blind to disturbances; feedback rejects disturbances but only after they appear. Combining them gives both anticipation and correction.

Why feedforward carries the load

A well-designed scenario provides most of the required actuation as feedforward, computed by the trajectory optimizer from a model. Feedback then handles only the residual error. This keeps feedback signals small, which keeps actuators away from saturation and makes the closed loop more stable and predictable.

The cost of leaning on feedback

If feedforward is poor, feedback must do heavy lifting, driving large corrections that can saturate actuators or excite oscillations. A control system that only reacts is always chasing; one that also anticipates stays ahead. This is why scenario optimization and feedback design are done together.

Model dependence

Feedforward is only as good as the model used to compute it. When the model is uncertain, feedforward is made conservative and feedback given more authority. As models improve, more of the burden shifts to feedforward. This balance is revisited as a machine's behavior becomes better understood through operation.

In the Kronos program

For the Hyperion breeder, feedforward waveforms come from trajectory optimization against validated models, and feedback loops trim the residuals. Because the machine is in simulation ahead of construction, feedforward is initially conservative; as operating data arrives after first plasma, the models sharpen and feedforward takes on more of the work through the between-shot loop.