Advanced AI and Robots From Zero to Building Your Own

Build your own image-sorting AI

Use Teachable Machine, a free website that runs in your browser, to teach an AI to tell two or three things apart using your computer’s camera, such as a pencil, a rubber and an empty hand. This is the same thing you did on the cat-and-dog board in Lesson 15, but with real pictures.

1. Ask an adult to open teachablemachine.withgoogle.com and choose an image project.
2. Name the first class “Pencil” and the second “Rubber”, then add a third class called “Nothing”.
3. Press the camera button for each class and turn the object to lots of different angles. Collect at least 50 pictures per class: close up and far away, in bright light and in the dark.
4. Press the Train button and wait less than a minute. Don’t close the window while it is training.
5. Hold an object in front of the camera and watch the confidence bar for each class. Try holding something you never taught it, such as scissors, and see what the AI says.
6. Try an experiment to see bias: teach it only with yellow pencils, then hold up a black pencil. Can the AI still recognise it? If not, add more varied examples and train it again.

A robot that follows a black line

This is the first robot that robotics students all over the world build, and it uses everything from Levels 1 to 5. Two infrared sensors look at the floor. A black line reflects little light and a white floor reflects a lot, so the robot knows which way it has drifted.

Parts

Wiring

The grounds of all the boards must be connected together. If you forget, the board can’t control the motors. This is the most common mistake beginners make. Leave the jumpers on the driver board’s ENA/ENB pins as they came from the factory.

The program

// Line follower: two IR sensors straddle a black line
const int SL = 2, SR = 3;             // left and right sensors
const int L1 = 5, L2 = 6, R1 = 9, R2 = 10;  // motor driver inputs

void drive(int left, int right) {     // speeds from -255 to 255
  analogWrite(L1, left  > 0 ? left  : 0);
  analogWrite(L2, left  < 0 ? -left : 0);
  analogWrite(R1, right > 0 ? right : 0);
  analogWrite(R2, right < 0 ? -right : 0);
}

void setup() {
  pinMode(SL, INPUT);  pinMode(SR, INPUT);
  pinMode(L1, OUTPUT); pinMode(L2, OUTPUT);
  pinMode(R1, OUTPUT); pinMode(R2, OUTPUT);
}

void loop() {
  bool left  = digitalRead(SL) == HIGH;   // HIGH = sensor sees black
  bool right = digitalRead(SR) == HIGH;
  if (!left && !right)     drive(150, 150);  // line is between: go straight
  else if (left && !right) drive(0, 160);    // drifted right: turn left
  else if (!left && right) drive(160, 0);    // drifted left: turn right
  else                     drive(0, 0);      // both black: stop
}

1. Install the Arduino IDE on your computer, plug the USB cable into the board, and choose Arduino Uno as the board type.
2. Type or copy the program above and press the Upload button. While you do this, keep the wheels lifted off the ground.
3. Stick black tape in a curving loop on a large sheet of white paper. The line should be about 2 centimetres wide, with gentle curves and no sharp corners.
4. Turn the screw on each sensor until the sensor’s light is on over the white floor and off over the black line, or the other way round, depending on the model.
5. Place the robot so the line is between the two sensors, switch on the battery holder, and watch it go.

An obstacle-avoiding robot

Use the same car base from Project 2, but swap its eyes from floor sensors to an HC-SR04 ultrasonic sensor. It sends out a sound too high-pitched for our ears to hear, then times how long the echo takes to come back, just as bats find their way in the dark. Sound travels at about 343 metres per second, so we can work out the distance from the time.

// Obstacle avoider: keep drive() and the motor pins from project 2,
// then replace setup() and loop() with these
const int TRIG = 7, ECHO = 8;

long distanceCm() {
  digitalWrite(TRIG, LOW);  delayMicroseconds(2);
  digitalWrite(TRIG, HIGH); delayMicroseconds(10);   // send a short ping
  digitalWrite(TRIG, LOW);
  long us = pulseIn(ECHO, HIGH, 30000);              // wait for the echo
  if (us == 0) return 999;                          // nothing in range
  return us * 0.0343 / 2;                           // there and back, so halve it
}

void setup() {
  pinMode(TRIG, OUTPUT); pinMode(ECHO, INPUT);
  pinMode(L1, OUTPUT); pinMode(L2, OUTPUT);
  pinMode(R1, OUTPUT); pinMode(R2, OUTPUT);
}

void loop() {
  if (distanceCm() < 20) {       // something within 20 cm
    drive(0, 0);        delay(200);
    drive(-140, -140);  delay(300);   // back up
    drive(150, -150);   delay(350);   // spin to the right
  } else {
    drive(150, 150);                  // path is clear
  }
  delay(50);
}

1. Mount the sensor on the front of the car, with its two “eyes” facing straight ahead.
2. Before fitting the wheels, open the Serial Monitor window and print out the distanceCm value. Move your hand closer and further away to check the distance is right.
3. Upload the full program, put the robot in a room with cardboard boxes as walls, and watch it find its way.
4. Try changing the number 20 to 10 or 40 and see whether the robot becomes more or less careful.