Five short lessons on proportional, integral, and derivative control - fly a drone to a target height, then steer a car around a curved track.
Answer first
Use the guided lessons to change proportional, integral, and derivative gains while a simulated drone holds altitude or a car follows a line. The simulator exposes overshoot, settling, and tracking error so you can connect each term to visible behavior. The gains are educational starting points and do not transfer directly to a physical plant.
Kp alone pushes thrust toward the target height, proportional to how far away the drone is. Push it up and the climb gets faster - but watch STEADY-STATE ERR: gravity keeps pulling down, and a proportional controller always needs some remaining error to keep producing enough thrust to fight it. It never quite reaches the line.
How far past the target height the drone climbs before settling - driven by momentum (or accumulated I) carrying past the setpoint.
How long until the height stays within a small tolerance band of the target for a full second.
The remaining gap between the target height and the drone's height once it's no longer actively correcting.
Design guide
Technical content reviewed
Worked approach
Raise proportional gain until response becomes useful, add integral action only when persistent error matters, then add derivative damping while watching noise sensitivity and overshoot. Recheck every term after the plant or operating point changes.
Common decisions
Proportional gain strengthens correction for the current error. More gain can improve response, but excessive gain can create oscillation, overshoot, actuator saturation, or sensitivity to unmodeled dynamics.
Add integral action when a persistent disturbance leaves steady-state error. Limit it when actuators saturate, because accumulated error can cause windup and a slow recovery.
Derivative action responds to rapid change. Sensor noise contains rapid changes too, so practical controllers filter the derivative term and choose sample timing carefully.
Related build evidence
See an ESP32 and NRF24L01+ architecture built for bidirectional commands and telemetry. It demonstrates control-system integration without claiming the simulator's gains were used on the vehicle.
See the projectApply it to real hardware
Send the plant behavior, sample rate, sensors, actuator limits, target response, and any test logs. Control design starts with the system, not copied gains.
Discuss the system