PID tuning and computer control issues

PID tuning is the process of choosing the gains Kp, Ki and Kd so that the closed-loop system meets its requirements: fast rise time, small overshoot, short settling time and zero steady-state error, while remaining stable. Methods include manual tuning, Ziegler-Nichols and software auto-tuning.

Practical issues in computer based control

1. Sampling and delay: the loop is sampled every T; a long T or computation delay reduces the phase margin and can cause instability. T should be much smaller than the plant time constant.
2. Quantisation: the ADC and DAC have finite resolution, which causes errors and can produce limit cycles; the derivative term (especially with a large Kd or a short sampling period) amplifies this quantisation noise.
3. Sensor noise: the derivative term amplifies noise, so filtering (low-pass or averaging) is required.
4. Integral wind-up: when the actuator saturates, the integral keeps growing; use clamping or anti-wind-up.
5. Actuator limits: the output has minimum and maximum values (PWM 0-100 %, throttle limits), and also slew-rate limits.
6. Numerical issues: fixed-point arithmetic, overflow and rounding on small microcontrollers.
7. Real-time scheduling: the control task must run at a fixed rate with low jitter; interrupts and RTOS priorities affect timing.
8. Model changes and non-linearity: gains tuned at one operating point may not work at another (e.g. different loads).
9. Set-point changes: a sudden set-point step gives a derivative kick; apply D to the measurement instead of the error.