> This is the hub article in the series. Start here if you want the full system view: sensors, controllers, vision processing, and autonomous flight working together as one embedded platform. If you mainly want a smaller BLE sensor node, low-power wearable, or TinyML edge-device path, continue with [Seeed Studio XIAO nRF52840 Complete Guide — BLE IoT Projects in Practice](/blog/ai/2026-03-03-xiao-nrf52840-iot-guide.en).
Introduction
Buying a drone and flying it is a completely different experience from **building one yourself and flying it**. Hardware assembly, PID tuning, sensor fusion, autonomous flight algorithms — this project brings together everything in control engineering.
Part 1: Hardware Configuration
Quadcopter Parts List
Essential Components:
├── Frame: F450 (450mm diagonal, ideal for beginners) ~$12
├── Motors: 2212 920KV BLDC × 4 ~$18
├── ESC: SimonK 30A × 4 (Electronic Speed Controller) ~$15
├── Propellers: 1045 (10-inch) × 4 (CW 2 + CCW 2) ~$4
├── Battery: 3S 11.1V 2200mAh LiPo ~$15
├── Controller: Arduino Mega 2560 or STM32 ~$12
├── IMU Sensor: MPU6050 (Accelerometer + Gyroscope) ~$3
├── Barometer: BMP280 (Altitude hold) ~$3
├── GPS: Neo-6M (For autonomous flight) ~$8
├── Receiver: FlySky FS-iA6B (with transmitter) ~$23
└── Power Distribution Board (PDB) + Connectors ~$4
Total: ~$117 (1/5 the price of a DJI Mini 4!)
Optional Components:
├── Raspberry Pi 4 (Vision processing, autonomous flight) ~$45
├── Pi Camera V2 (Object tracking) ~$23
├── Ultrasonic Sensor: HC-SR04 (Low altitude measurement) ~$2
├── Optical Flow Sensor: PMW3901 (Indoor position hold) ~$12
└── Telemetry: HC-12 433MHz (Ground monitoring) ~$4
Motor Layout
Front
M1(CW) M2(CCW)
\ /
\ /
\ /
[FC Board]
/ \
/ \
/ \
M3(CCW) M4(CW)
Back
CW = Clockwise, CCW = Counter-Clockwise
Diagonal motors rotate in the same direction!
→ Torque cancellation for airframe stability
Wiring Diagram
Battery (3S LiPo 11.1V)
│
├──[PDB]── ESC1 ── M1
│ ESC2 ── M2
│ ESC3 ── M3
│ ESC4 ── M4
│
├──[BEC 5V]── Arduino Mega
│ ├── MPU6050 (I2C: SDA→D20, SCL→D21)
│ ├── BMP280 (I2C: same bus)
│ ├── GPS Neo-6M (Serial2: TX→D16, RX→D17)
│ ├── Receiver (PPM→D2 or individual channels)
│ └── ESC Signal (D3, D5, D6, D9)
│
└──[5V]── Raspberry Pi 4 (USB or Serial connection)
└── Pi Camera
Part 2: PID Control — The Heart of the Drone
What is PID?
Goal: Keep the drone level (Roll = 0°)
Current state: Roll = 5° (tilted to the right)
Error = Setpoint - Current = 0° - 5° = -5°
PID Output = P + I + D
P (Proportional): Correction proportional to the error
→ -5° × Kp = immediate response, but may oscillate
I (Integral): Corrects accumulated error
→ Eliminates steady-state offset, but too much causes overshoot
D (Derivative): Corrects the rate of change of error
→ Suppresses sudden changes, prevents oscillation
Arduino PID Implementation
// PID Controller Class
class PIDController {
private:
float Kp, Ki, Kd;
float prevError = 0;
float integral = 0;
float maxIntegral = 300; // Prevent integral windup
unsigned long prevTime = 0;
public:
PIDController(float p, float i, float d)
: Kp(p), Ki(i), Kd(d) {}
float compute(float setpoint, float measured) {
unsigned long now = micros();
float dt = (now - prevTime) / 1000000.0f; // In seconds
if (dt <= 0 || dt > 0.5) dt = 0.004; // Safety fallback
prevTime = now;
float error = setpoint - measured;
// P: Proportional
float P = Kp * error;
// I: Integral (with windup prevention)
integral += error * dt;
integral = constrain(integral, -maxIntegral, maxIntegral);
float I = Ki * integral;
// D: Derivative (based on measurement — prevents kick on setpoint change)
float derivative = (error - prevError) / dt;
float D = Kd * derivative;
prevError = error;
return P + I + D;
}
void reset() {
prevError = 0;
integral = 0;
}
};
// Separate PID for Roll, Pitch, and Yaw
PIDController rollPID(1.2, 0.04, 18.0);
PIDController pitchPID(1.2, 0.04, 18.0);
PIDController yawPID(3.0, 0.02, 0.0);
Sensor Fusion (Complementary Filter)
#include <Wire.h>
#include <MPU6050.h>
MPU6050 mpu;
float roll = 0, pitch = 0, yaw = 0;
const float ALPHA = 0.98; // Complementary filter coefficient
void updateIMU() {
int16_t ax, ay, az, gx, gy, gz;
mpu.getMotion6(&ax, &ay, &az, &gx, &gy, &gz);
// Accelerometer → Absolute angle (slow but no drift)
float accelRoll = atan2(ay, az) * 180.0 / PI;
float accelPitch = atan2(-ax, sqrt(ay*ay + az*az)) * 180.0 / PI;
// Gyroscope → Angular velocity integration (fast but drifts)
float dt = 0.004; // 250Hz
float gyroRollRate = gx / 131.0; // °/s
float gyroPitchRate = gy / 131.0;
float gyroYawRate = gz / 131.0;
// Complementary filter: Gyro (short-term) + Accelerometer (long-term) fusion
roll = ALPHA * (roll + gyroRollRate * dt) + (1 - ALPHA) * accelRoll;
pitch = ALPHA * (pitch + gyroPitchRate * dt) + (1 - ALPHA) * accelPitch;
yaw += gyroYawRate * dt; // Gyro only (yaw cannot be derived from accelerometer)
}
Motor Mixing
#include <Servo.h>
Servo motor[4];
void setup() {
motor[0].attach(3); // Front-left (CW)
motor[1].attach(5); // Front-right (CCW)
motor[2].attach(6); // Rear-left (CCW)
motor[3].attach(9); // Rear-right (CW)
// ESC initialization (1000~2000μs PWM)
for (int i = 0; i < 4; i++) {
motor[i].writeMicroseconds(1000);
}
delay(2000);
}
void setMotors(float throttle, float rollOut, float pitchOut, float yawOut) {
// Motor mixing formula
float m1 = throttle + pitchOut + rollOut - yawOut; // Front-left CW
float m2 = throttle + pitchOut - rollOut + yawOut; // Front-right CCW
float m3 = throttle - pitchOut + rollOut + yawOut; // Rear-left CCW
float m4 = throttle - pitchOut - rollOut - yawOut; // Rear-right CW
// Range limiting (1000~2000μs)
motor[0].writeMicroseconds(constrain(m1, 1100, 1900));
motor[1].writeMicroseconds(constrain(m2, 1100, 1900));
motor[2].writeMicroseconds(constrain(m3, 1100, 1900));
motor[3].writeMicroseconds(constrain(m4, 1100, 1900));
}
Main Loop (250Hz)
void loop() {
// 1. Read sensors
updateIMU();
// 2. Read controller input
float targetRoll = map(rcChannel[0], 1000, 2000, -30, 30);
float targetPitch = map(rcChannel[1], 1000, 2000, -30, 30);
float targetYaw = map(rcChannel[3], 1000, 2000, -180, 180);
float throttle = rcChannel[2];
// 3. PID computation
float rollOut = rollPID.compute(targetRoll, roll);
float pitchOut = pitchPID.compute(targetPitch, pitch);
float yawOut = yawPID.compute(targetYaw, yaw);
// 4. Motor output
if (throttle > 1100) { // Safety: only when throttle exceeds minimum
setMotors(throttle, rollOut, pitchOut, yawOut);
} else {
for (int i = 0; i < 4; i++)
motor[i].writeMicroseconds(1000); // Stop motors
}
// Maintain 250Hz
while (micros() - loopTimer < 4000);
loopTimer = micros();
}
Part 3: Raspberry Pi Vision Control
Object tracking on Raspberry Pi + sending commands to Arduino
Serial communication with Arduino
arduino = serial.Serial('/dev/ttyACM0', 115200)
Camera
cap = cv2.VideoCapture(0)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 320)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 240)
Color tracking (follow a red object)
while True:
ret, frame = cap.read()
hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
Red color mask
mask = cv2.inRange(hsv, (0, 120, 70), (10, 255, 255))
contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
if contours:
largest = max(contours, key=cv2.contourArea)
M = cv2.moments(largest)
if M["m00"] > 500:
cx = int(M["m10"] / M["m00"])
cy = int(M["m01"] / M["m00"])
Error from screen center (160,120)
error_x = cx - 160 # Horizontal
error_y = cy - 120 # Vertical
Send correction command to Arduino
cmd = struct.pack('hh', error_x, error_y)
arduino.write(cmd)
Part 4: Practical Control System Examples
Inverted Pendulum
// Inverted pendulum = scaled-down version of drone control
// Balancing a rod vertically = keeping a drone level
// Read angle with rotary encoder
volatile long encoderCount = 0;
float pendulumAngle = 0;
void encoderISR() {
encoderCount += (digitalRead(ENCODER_B) == HIGH) ? 1 : -1;
}
void setup() {
attachInterrupt(digitalPinToInterrupt(ENCODER_A), encoderISR, RISING);
}
void loop() {
pendulumAngle = encoderCount * 360.0 / 2400.0; // PPR=600, x4
// Motor control with PID
float output = balancePID.compute(0, pendulumAngle); // Target: 0°
setMotor(output);
}
Line Tracer (Autonomous Driving Basics)
// Line detection with 5 IR sensors
int sensors[5] = {A0, A1, A2, A3, A4};
float getLinePosition() {
int values[5];
float weighted = 0, total = 0;
for (int i = 0; i < 5; i++) {
values[i] = analogRead(sensors[i]);
weighted += values[i] * (i - 2) * 1000; // -2000 ~ +2000
total += values[i];
}
return weighted / total; // -2000(left) ~ +2000(right)
}
void loop() {
float position = getLinePosition(); // Current position
float correction = linePID.compute(0, position); // Target: center (0)
int leftSpeed = baseSpeed + correction;
int rightSpeed = baseSpeed - correction;
setMotors(leftSpeed, rightSpeed);
}
PID Tuning Guide
Ziegler-Nichols Method:
1. Start with Ki = 0, Kd = 0
2. Gradually increase Kp until oscillation begins — find this value (Ku)
3. Measure the oscillation period (Tu)
4. Calculate:
Kp = 0.6 × Ku
Ki = 2 × Kp / Tu
Kd = Kp × Tu / 8
Practical Tips:
├── Start with P only: Fast response, slight oscillation is OK
├── Add D: Dampen oscillation (10~20x of P)
├── Add I: Eliminate steady-state error (keep it small!)
└── For drones, too much I is dangerous (integral windup)
Read Next in This Embedded Hands-On Series
- [Seeed Studio XIAO nRF52840 Complete Guide — BLE IoT Projects in Practice](/blog/ai/2026-03-03-xiao-nrf52840-iot-guide.en)
Read this next if you want to move from a large control system to a compact BLE sensor node, low-power wearable, or TinyML edge-device project.
**Q1.** What are the roles of P, I, and D in PID?
||P (Proportional): Immediate correction proportional to error. I (Integral): Eliminates accumulated error (steady-state error). D (Derivative): Suppresses sudden changes (prevents oscillation)||
**Q2.** Why do diagonal motors on a quadcopter rotate in the same direction?
||Torque cancellation. CW and CCW motors are placed diagonally so the total torque is zero. With same-direction motors only, the airframe would spin.||
**Q3.** What does ALPHA = 0.98 mean in the complementary filter?
||Fusion at 98% gyro (short-term, fast) + 2% accelerometer (long-term, stable). The gyro is fast but drifts; the accelerometer is slow but provides absolute values.||
**Q4.** What is Integral Windup?
||A phenomenon where the integral value keeps growing while motor output is saturated. Causes excessive overshoot when saturation is released. Prevented by constraining the integral value.||
**Q5.** What do 1000~2000μs mean in the ESC PWM range?
||1000μs: Motor stop. 2000μs: Maximum rotation. 1500μs: Mid-range. The ESC receives this PWM signal and controls the 3-phase current of the BLDC motor.||
현재 단락 (1/244)
Buying a drone and flying it is a completely different experience from **building one yourself and f...