Python PID Control Systems — Core Concepts

The Control Problem

A control system takes a desired value (setpoint), measures the actual value (process variable), and adjusts an output to minimize the difference (error). A thermostat set to 22°C measures the room at 19°C, sees a 3°C error, and turns on the heater.

PID is the most widely used control algorithm in industry. An estimated 95% of industrial control loops use some form of PID. Its popularity comes from being simple to implement, working well across diverse systems, and requiring only three parameters to tune.

The Three Terms

Proportional (P)

The proportional term produces output proportional to the current error:

Output_P = Kp × error

If Kp is 2 and the error is 5, the output is 10. Larger errors produce stronger corrections. The problem: P alone usually cannot eliminate error completely. A system with only P control settles at a steady-state error — close to the target but not quite there. Increasing Kp reduces this offset but makes the system oscillate more.

Integral (I)

The integral term accumulates error over time:

Output_I = Ki × ∑(error × dt)

Even a tiny persistent error will accumulate and eventually drive the integral term large enough to correct it. This eliminates steady-state error. The risk: if the error persists for a long time (during startup or a large setpoint change), the integral can grow enormous — a problem called “integral windup” that causes severe overshoot.

Derivative (D)

The derivative term responds to the rate of change of error:

Output_D = Kd × (Δerror / Δt)

When the error is shrinking rapidly (the system is approaching the setpoint fast), D applies a braking force. This dampens oscillation and reduces overshoot. The downside: D amplifies noise. If the sensor reading jitters, the derivative sees rapid changes and produces noisy output. Filtering the derivative input or the measurement signal is standard practice.

Combined PID Output

Output = Kp × e + Ki × ∫e dt + Kd × de/dt

The three gains (Kp, Ki, Kd) determine behavior:

Increase	Rise Time	Overshoot	Settling Time	Steady-State Error
Kp	Decreases	Increases	Small change	Decreases
Ki	Decreases	Increases	Increases	Eliminates
Kd	Small change	Decreases	Decreases	No effect

Tuning Methods

Manual Tuning

1. Set Ki and Kd to zero
2. Increase Kp until the system oscillates consistently
3. Add Kd to dampen the oscillation
4. Add Ki to eliminate remaining offset
5. Fine-tune all three iteratively

Ziegler-Nichols Method

A structured approach:

1. Set Ki = 0, Kd = 0
2. Increase Kp until the system sustains steady oscillations (the “ultimate gain” Ku)
3. Measure the oscillation period (Tu)
4. Apply these formulas:

Controller	Kp	Ki	Kd
P only	0.50 × Ku	—	—
PI	0.45 × Ku	0.54 × Ku / Tu	—
PID	0.60 × Ku	1.2 × Ku / Tu	0.075 × Ku × Tu

Ziegler-Nichols gives aggressive tuning (significant overshoot). Most practitioners start with these values and back off for smoother response.

Real-World Applications

• Drone altitude hold: P corrects altitude error, I compensates for wind and weight changes, D prevents bouncing
• Line-following robot: P steers proportional to deviation from the line, D prevents weaving
• 3D printer heated bed: PI control (D usually not needed for slow thermal systems)
• Autonomous vehicle speed control: PID manages throttle/brake to maintain target speed

Common Misconception

“You always need all three terms.” Many real systems work perfectly with just PI control (no derivative). Thermal systems, liquid level control, and pressure regulation often skip the D term because the process is naturally slow and noise-free. Adding D to a noisy system without filtering makes things worse. Start with P, add I if you need zero steady-state error, and only add D if you have oscillation problems.

One thing to remember: Tuning a PID controller means finding the right balance of three knobs — proportional for speed, integral for accuracy, derivative for smoothness — and Python lets you simulate the system’s response before deploying to real hardware.