English is currently available.
Introduction DISCUSSION
For someone beginning control-system projects, what small experiment best connects the equations to real measurements? A low-voltage fan-speed rig with a tachometer seems more manageable than a robot that combines steering, balance and navigation.
Which signals should be logged, and what should be understood before adding integral and derivative terms to the controller?
First measure the open-loop response. Apply a bounded change in drive command and log command, measured speed and time. Look for startup thresholds, saturation, delay and noise. The drive command is not the same as motor speed, and low duty cycles may not overcome friction. Keep the mechanism guarded and use a suitable motor driver and supply for the experiment.
Use a fixed sample interval and start with proportional control. Log the setpoint, error and output together so oscillation and steady-state error are visible. If the loop runs every 10 ms, its update rate is 100 Hz. That does not imply it can control a 100 Hz plant; sampling must be selected relative to the dynamics and delays you actually measured.
When adding integral action, account for output limits. Otherwise the integrator can accumulate error while the motor command is already saturated, causing a long recovery. Derivative action is sensitive to measurement noise and often needs filtering. Change one term at a time and compare response plots using the same setpoint change, rather than declaring improvement from a single visually smooth run.
