Spark Academy — Learn Electrical Engineering by Building

The mission

Build the classic first robot: two driven wheels, two downward-looking reflectance sensors, and firmware that steers so the black line stays between them. It is Lesson 14.2’s PID with wheels bolted on — and the moment it takes its first corner on its own, you will understand why people never stop building robots.

Shopping list (beyond Pico + breadboard kit)

Part	Notes	≈ Cost
2WD robot chassis kit	two yellow “TT” gear motors, wheels, caster, plate — the universal starter chassis	$12–18
DRV8833 driver module	dual H-bridge with interlocks; happier at 3.3 V logic than the old L298N	$3
2× TCRT5000 reflectance modules	IR LED + phototransistor — a divider output per sensor (2.2 again)	$3
4×AA battery holder	motor power — separate from the Pico’s USB/5V supply	$3
Black electrical tape	the racetrack. Light floor, wide gentle curves to start	$2

The wiring

#	From	To	Notes
1	Battery pack + / −	DRV8833 VM / GND	motor power stays off the Pico’s rail
2	Pico GND	DRV8833 GND & sensor GNDs	common ground — the scope capstone’s oldest law
3	GP2, GP3	DRV8833 AIN1, AIN2	left motor (PWM both for fwd/rev)
4	GP4, GP5	DRV8833 BIN1, BIN2	right motor
5	DRV8833 AOUT/BOUT	the two motors	swap a motor’s two wires if it runs backwards
6	Sensor AO (left / right)	GP26 / GP27	the night-light’s ADC pins, now looking at the floor
7	Sensor VCC	Pico 3V3	sensors sip; motors must not touch this rail
8	Sensors, mounted	~15 mm apart, 3–8 mm above floor, ahead of the axle	height matters more than you think

The firmware

from machine import Pin, PWM, ADC
from time import sleep_ms

# DRV8833: two PWM pins per motor (fwd / rev)
m_l_f, m_l_r = PWM(Pin(2)), PWM(Pin(3))
m_r_f, m_r_r = PWM(Pin(4)), PWM(Pin(5))
for m in (m_l_f, m_l_r, m_r_f, m_r_r):
    m.freq(1000)

sens_l = ADC(26)   # left line sensor
sens_r = ADC(27)   # right line sensor

KP, KD = 28000, 900
BASE = 28000       # of 65535
last_err = 0

def drive(m_f, m_r, v):
    v = max(-65535, min(65535, int(v)))
    m_f.duty_u16(v if v > 0 else 0)
    m_r.duty_u16(-v if v < 0 else 0)

while True:
    l = sens_l.read_u16() / 65535
    r = sens_r.read_u16() / 65535
    err = r - l                    # +: line is to the right
    steer = KP * err + KD * (err - last_err)
    last_err = err
    drive(m_l_f, m_l_r, BASE + steer)   # line right -> left wheel faster
    drive(m_r_f, m_r_r, BASE - steer)   # -> robot turns right, toward the line
    sleep_ms(10)                   # 100 decisions per second

Thirty lines. The loop is the night-light’s superloop; the sensors are dividers into ADCs; the steering is 14.2’s PD law; the motors are Unit 8’s PWM through 19.1’s H-bridge. Tune KP exactly as the simulator taught you: too low corners wide, too high snakes down the straights, KD calms the snake. Then raise BASE until bravery fails.

If it misbehaves

Symptom	Likely cause	Fix
Turns away from the line	sensors or steering sign swapped	swap GP26/GP27, or negate err — one change (15.2!)
One motor runs backwards	motor leads reversed	swap that motor’s two wires at the driver
Pico resets when motors kick	motor noise sagging the supply	separate battery pack for VM, common grounds, 100 µF across VM (15.1’s decoupling!)
Blind to the line	sensor height / ambient IR	3–8 mm off the floor, shade from sunlight, check readings over tape vs floor in Thonny
Snakes violently	your Kp is shouting	halve KP, add KD — you rehearsed this exact tuning in the twin

Where this road leads

Add encoders (19.3) and it knows its speed. Add an IMU and it can balance. Swap the tape for walls and a distance sensor: a maze solver. Everything else in robotics — SLAM, manipulators, autonomy — is more sensors, more loops, more math on the same skeleton you just built: sense → decide → act, at loop rates you now know how to choose, filter (13.3) and tune (14.2). Welcome to robotics. You came the long way, and it shows.

Capstone: Build the Line Follower