Upgrade to Pro — share decks privately, control downloads, hide ads and more …

How I built a Neural Network controlled self dr...

Avatar for dps dps
April 21, 2012

How I built a Neural Network controlled self driving (RC) Car and you can too

Cheap, mass produced mobile phone hardware laden with sensors combined with the compute power of the machines we use every day make expensive research systems of the 90s implementable in your living room. I'll explain how I did just that. Combining machine learning techniques anyone can learn online, an open source hardware platform, a simple mobile app and a bit of heath-robinson device hacking to build a remote control car that learns how to navigate autonomously.

Avatar for dps

dps

April 21, 2012
Tweet

Other Decks in Programming

Transcript

  1. void setup() { Serial.begin(9600); pinMode(forwardPin, OUTPUT); ... } void loop()

    { ... if (Serial.available() > 0) { incomingByte = Serial.read(); left = right = forward = back = LOW; if (incomingByte & 0x01) { left = HIGH; } if (incomingByte & 0x02) { right = HIGH; } if (incomingByte & 0x04) { forward = HIGH; } if (incomingByte & 0x08) { back = HIGH; } ... digitalWrite(leftPin, left); digitalWrite(rightPin, right); digitalWrite(forwardPin, forward); digitalWrite(backPin, back); digitalWrite(ledPin, error); } } Saturday, 21 April 12
  2. Training • The pipeline • Feature Normalization • Regularization •

    Training time + PCA? • Accuracy on training data Saturday, 21 April 12