The Basics Behind AI Models for Self-Driving Cars | by Claudia Ng

Learn to build a neutral network that can drive using PyTorch in Python

En route via robo-taxi (Image by author)

I recently took my first robo-taxi ride in San Francisco.

I got in the back and the driver’s seat was empty.

I watched in awe as the car signaled to do a right turn at the stop sign and waited for pedestrians to cross. It then slowly accelerated and the steering wheel turned right.

It was such a smooth ride. No more worrying about putting my life in the hands of a sleepy or cranky taxi driver.

As a Data Scientist, I am fascinated with the technology powering autonomous vehicles. So, I learned to build a simple neural network that can predict how to drive that I’ll walk you through.

We need to first understand the mechanics of how the software and hardware components fit together.

A car moves on a horizontal plane and can go in four different directions. So, the car is outfitted with sensors to detect its proximity to objects in 4 directions: