Actuator Saturation & Anti-Windup

Problem Setup

Spacecraft Attitude Control with a Torque Limit

Consider a simplified single-axis spacecraft attitude model. The controller commands a torque, but the real actuator can only deliver torque within a finite physical range.

$$ \dot{\theta} = \omega $$ $$ \dot{\omega} = \tau $$

The physical actuator constraint is:

$$ |\tau| \leq \tau_{max} $$

Controllers assume unlimited authority.
Real spacecraft do not.

PD Limitation

Why PD Control Can Hit Physical Limits

A PD controller uses attitude error and angular velocity:

$$ \tau_c = K_p(\theta_{target} - \theta) - K_d\omega $$

A high $K_p$ can make the response faster in an ideal model, but it can also demand torque that the actuator cannot deliver. Once the actuator clips the command, the actual spacecraft response no longer follows the ideal controller design.

High gain does not guarantee high authority. If the actuator saturates, the applied torque is limited.

Integral Action

Why PID Adds a New Risk

PID adds accumulated error:

$$ \tau_c = K_p e + K_i \int e\,dt - K_d\omega $$

The integral term is useful because it can remove steady-state error caused by bias or disturbance. But when the actuator is saturated, the integral term may keep growing even though the actuator cannot deliver the requested torque.

$$ I_{k+1} = I_k + e_k\Delta t $$

Actuator Saturation

What Saturation Means

Saturation means the controller asks for one torque, but the actuator applies a clipped torque.

$$ \tau_{applied} = \text{clip}(\tau_c, -\tau_{max}, \tau_{max}) $$

In real spacecraft control, this difference is critical:

commanded torque ≠ applied torque

The torque plot in the interactive demo makes this visible by showing both the commanded torque and the clipped applied torque.

Integrator Windup

When Integral Action Stores Too Much Error

If the actuator is already saturated, adding more integral action cannot produce more applied torque. The integral term may continue growing, storing excess control demand.

When the spacecraft finally approaches the target, this stored integral energy can cause overshoot, oscillation, slow recovery, or degraded pointing behaviour.

PID without anti-windup:
integral grows $ \rightarrow $ overshoot and delayed recovery

Anti-Windup

Do Not Integrate While Saturated

A simple anti-windup method is conditional integration: stop accumulating the integral term while the actuator is saturated.

if saturated $ \rightarrow $ stop integral growth

if (!isSaturated) {
  integralError += error * dt;
}

This does not make the actuator stronger. It simply prevents the controller from building an unrealistic internal command that the actuator cannot physically apply.

Interactive Demo

Compare PD, PID, and PID with Anti-Windup

Select a controller mode and adjust the gains. The simulation includes disturbance torque and actuator saturation. Watch the attitude response, pointing error, torque clipping, integral growth, and engineering metrics update automatically.

Controller mode:

PD PID PID + Anti-windup

Kp: 6

Ki: 1.0

Kd: 2

Torque limit: 1.0

Disturbance torque: 0.25

Adjust any slider or select a controller mode to run the simulation.

Compare All Modes

This comparison uses the same gains, disturbance, and torque limit for PD, PID, and PID with anti-windup. The key visual moment is the difference between uncontrolled integral growth and controlled integral behaviour.

Watch this carefully: PID without anti-windup shows large overshoot due to stored integral error, while anti-windup produces a cleaner response.

Engineering insights will appear after the simulation runs.

Engineering Metrics

Automatic Control Diagnostics

The page computes practical diagnostic metrics that matter in real actuator-limited systems.

Saturation Time

Waiting for simulation...

Windup Detector

Waiting for simulation...

Recovery / Settling Time

Waiting for simulation...

Overshoot

Waiting for simulation...

Max Commanded Torque

Waiting for simulation...

Takeaway

What This Problem Shows

PD

Can be stable, but high gains may demand torque beyond actuator authority.

PID

Can reduce bias, but saturation can cause integral windup and overshoot.

Anti-Windup

Prevents unrealistic integral growth and improves recovery under torque limits.

The final engineering lesson is:

Control design is not just math.
It must respect physical limits.

Good control design = stability + performance + physical feasibility