Feedforward vs Feedback

Why a controller can predict the needed command, correct the result afterward, and still mislead diagnosis when those two jobs are treated as the same thing.

Conceptual definition

Feedforward and feedback are two different ways a controller decides what command to send.

Feedforward control predicts the command that should be needed from known operating inputs, calibration tables, and an internal model of the system. It acts before the final result has been measured.

Feedback control compares the achieved result with the desired result and applies correction after an error is detected.

Most automotive control systems blend both. A base command is established from load, speed, temperature, pressure, driver request, or other inputs. Feedback then trims that command when the actual result does not match the target.

Why it varies

The balance between feedforward and feedback changes with the system, operating state, and speed of the event.

Fast transients often depend heavily on feedforward because the system cannot wait for a delayed feedback signal. Steady operation may allow feedback to make smaller, more accurate corrections.

Feedback authority may also be reduced, delayed, frozen, or disabled during startup, warm-up, deceleration, heavy load, protection modes, sensor faults, or other strategy states.

What it is not

Feedforward is not simply another name for open-loop operation, and feedback is not automatically the same as one global closed-loop state.

A controller can use feedback in one part of the system while another part is operating from modeled or fixed commands.

A large feedback correction is not proof that the feedback sensor is defective. It proves only that the controller sees a difference between the target and the reported result.

No visible correction is not proof that the system is accurate. Correction may be disabled, limited, delayed, learned into another value, or already pinned at an authority limit.

Failure modes

Feedforward and feedback faults can produce similar symptoms while originating in different parts of the control chain.

The common diagnostic mistake is to blame the signal that reports the mismatch without proving whether the mismatch was created by the base calculation, the command path, the actuator, the mechanical system, or the feedback measurement itself.

How SpeedNeeds uses it

SpeedNeeds separates the control chain into target, feedforward command, feedback correction, final command, and achieved result.

Diagnosis should establish what the controller wanted, what base command it predicted, what correction it added, whether the actuator received and delivered that command, and whether the measured result was valid.

Data should be read in time order. A correction that appears after the result drifts is different evidence from a correction that creates the drift.

When correction is large, the next move is not automatic sensor replacement. The next move is to prove the target, base inputs, actuator response, physical delivery, feedback accuracy, and authority limits under the condition where the error occurs.

Caution and edge cases

Scan-tool labels do not always expose the strategy cleanly. A displayed command may already include adaptive values, limits, protection logic, and feedback correction.

Some controllers switch between control methods without presenting an obvious mode change. Others retain learned corrections across key cycles or clear them after battery disconnect, reflash, or reset.

A zeroed or frozen correction PID may mean the controller is satisfied, but it may also mean feedback is unavailable, disabled, filtered, or not represented by that PID.

Always verify how the specific system defines target, base command, correction, and achieved value before interpreting the data.

Closing clarity

Feedforward explains why the controller started with a command. Feedback explains why it changed that command after seeing the result.

The correction is evidence of a mismatch, not proof of which part caused it. Strong diagnosis follows the entire chain from predicted command to actual response before assigning blame.