Servo control loops react to error: the motor lags the command, the controller sees the lag, and it pushes harder. Feedforward skips the waiting: it predicts the effort the move will require and applies it proactively, so the feedback loop only corrects the residual. For high acceleration servos, where the inertial forces dominate and the lag without feedforward is severe, feedforward gain tuning is the difference between a servo that tracks crisply and one that perpetually chases its command. The gains must match the mechanism's physics, and the tuning procedure finds that match.
What Feedforward Compensates
The dominant feedforward terms mirror the mechanism's dynamics. Velocity feedforward compensates viscous friction and back-EMF: the effort proportional to speed. Acceleration feedforward compensates inertia: the effort proportional to acceleration, which is mass times acceleration at its simplest. For high acceleration moves, the acceleration term dominates; the servo needs a large proactive current spike to get the mass moving, and feedback alone always delivers it late.
Without feedforward, the feedback controller must generate all of this effort from observed error, which means error must exist before effort appears: the servo lags during acceleration, overshoots at the end, and the tracking error traces the acceleration profile. Feedforward supplies the predictable effort in advance, leaving feedback to handle only the unpredictable: friction variations, disturbances, and model imperfections.
Tuning the Acceleration Gain
The acceleration feedforward gain should equal the system's inertia as seen by the controller: the gain that converts commanded acceleration into the exact current (or torque command) the inertia requires. The tuning procedure is empirical: command a brisk acceleration move with the gain at zero and observe the tracking error during the acceleration phase; increase the gain and repeat. As the gain approaches the true inertia value, the acceleration-phase tracking error shrinks toward zero.
The signature of correct tuning is flat tracking error through the acceleration: the servo neither lags (gain too low) nor leads (gain too high). Overshooting the gain produces a leading error, the servo jumping ahead of the command during acceleration, which is just as bad as lagging; the goal is the null, not the maximum.
Tuning the Velocity Gain
Velocity feedforward compensates the speed-proportional losses. Tune it with constant-velocity moves: command a steady slew and observe the steady-state following error; increase the velocity gain until that error nulls. The velocity gain is typically tuned after or alongside the acceleration gain, since real moves combine both phases and the gains interact slightly at the transitions.
- Tune one axis at a time with the others stationary; coupled dynamics confuse single-axis tuning.
- Use moves representative of the actual application; tuning on gentle moves under-tunes for aggressive ones.
- Verify at multiple speeds and accelerations; a gain correct at one operating point may err at another.
- Re-tune after any mechanical change; inertia changes invalidate the acceleration gain directly.
Stability and the Limits of Feedforward
Feedforward does not affect stability the way feedback gains do: it is an open-loop injection that does not participate in the loop's dynamics, so increasing feedforward gains does not cause oscillation. This makes feedforward the safe direction for performance improvement when the feedback gains are already at their stability limits. The practical limit is model accuracy: feedforward beyond the true physical values injects error rather than canceling it.
Friction's nonlinearities, stiction, Coulomb friction, and position-dependent variations, cannot be fully captured by linear feedforward terms. The residual errors from these effects remain for the feedback loop; feedforward handles the big predictable physics, feedback handles the messy reality. Expect improvement, not perfection.
Validating the Tuned System
The validation is the application's actual motion profile with tracking error recorded: the error during acceleration, during constant velocity, and at the move's end should all be small and, critically, consistent across repetitions. Inconsistent errors point to unmodeled effects (varying friction, backlash, compliance) that no feedforward tuning will fix; consistent errors point to gain adjustments still available.
Document the final gains with the mechanism configuration and the test moves used. Feedforward gains are married to the inertia they compensate; any future mechanical change, added payload, different end effector, starts the tuning over. The documentation makes that restart quick instead of mysterious.
Null the acceleration error with the inertia gain, null the velocity error with the friction gain, validate on real moves, and document everything. Feedforward is the controller's knowledge of physics made actionable; tuned well, the servo stops chasing and starts arriving.
Related Guides
- Anti-Windup Reset Tuning in PID Loops: Integral windup is the PID loop's original sin: while the actuator is saturated, the integrator keeps accumulating error, building up a huge
- Complementary Filter Tuning for Six-Axis IMU Drift: A six-axis IMU gives you a gyroscope and an accelerometer, and neither one tells the truth by itself. The gyroscope measures rotation rate
- Extended Kalman Filter Fusion for Robot IMU Drift: An inertial measurement unit, the accelerometer plus gyroscope chip in every robot, tells you how the robot is moving right now, but integrating
- Low-Pass Noise Filtering on Raw Gyroscope Inputs: A raw gyroscope signal looks nothing like the smooth rotation it represents. Zoom in and the trace is a thick band of jitter: vibration from motors