Line Follower & PID Tuning

A line-following robot uses one or more light sensors (usually IR reflectance sensors) to detect a contrasting line on the floor and steer itself to stay centered on it. The raw sensor reading only tells you how far off-center you are — the error. A PID controller (Proportional-Integral-Derivative) turns that error into a smooth, stable steering correction so the robot doesn't zig-zag off the track or overshoot every curve.

The PID control law computes a correction $u(t)$ from the error $e(t)$ (deviation from the line center):

$u(t) = K_p\,e(t) + K_i\int_0^t e(\tau)\,d\tau + K_d\,\frac{de(t)}{dt}$

In discrete robot code, this is usually written as:

error      = setpoint - sensorReading
integral   = integral + error
derivative = error - lastError
correction = Kp*error + Ki*integral + Kd*derivative
lastError  = error

The correction is added to one motor's base speed and subtracted from the other, steering the robot back onto the line.

1. Read the sensor array and compute a weighted position of the line (e.g. average of sensor indices weighted by how dark each one reads).
2. Compute error = center - linePosition.
3. Start tuning with only $K_p$ (set $K_i = K_d = 0$). Increase $K_p$ until the robot follows the line but starts to oscillate (wobble side to side).
4. Add $K_d$ to damp that oscillation — it "predicts" the error and slows the correction as the robot approaches the line, reducing overshoot.
5. Add a small $K_i$ only if the robot settles slightly off-center on straight sections (steady-state error) — too much $K_i$ causes slow, sweeping oscillations.
6. Re-test on both sharp corners and straight sections, adjusting all three gains until motion is smooth at the robot's top speed.

Tuning symptoms cheat-sheet:

• Sluggish, cuts corners → increase $K_p$.
• Fast oscillation / wobble → increase $K_d$ or decrease $K_p$.
• Slow drift never fully corrected → increase $K_i$ slightly.
• Slow, wide oscillation → decrease $K_i$.

Line followers are the simplest working example of closed-loop feedback control, the same principle behind cruise control, drone stabilization, thermostats, and industrial robot arms holding a position. Warehouse AGVs (automated guided vehicles) at companies like Amazon use an evolved version of line/path following with PID or more advanced controllers to move shelves and pallets precisely along fixed paths.

1. Your robot follows the line but wobbles rapidly side to side even on straight sections. Which gain should you adjust first, and in which direction?
2. Why is the derivative term useful for reducing overshoot, when it only looks at how fast the error is changing rather than the error itself?