The self-driving car

After years of development, Google’s self-driving car project enters its next phase. Previously, the company has added its driverless technology to existing vehicles – effectively turning ordinary cars into self-driving prototypes. Now they’ve unveiled their own model, which they hope to have on the road as early as next year.

Google’s self-driving car is a two-person vehicle that can travel up to 40 kilometres an hour. Inside, the driverless pod is quite different from a conventional car: there’s no steering wheel, accelerator or brake pedal, just a single button to start and stop. On the outside, safety measures include a soft front panel and flexible windscreen – both intended to minimise injuries to pedestrians in the event of a collision.

The car collects data about its environment using a combination of lasers, radar sensors, and cameras. These are analysed and interpreted by powerful onboard computers, allowing the car to navigate through traffic safely and efficiently. Key to the vehicle’s ability to avoid obstacles is an 18^th century maths theorem known as Bayes’ theorem.

At its simplest, Bayes’ theorem can be expressed as: hypothesis + new objective data = improved hypothesis. Applied to robotics, this allows a system to calculate probability and make reliable predictions, thereby determining what course of action to take based on the situation. This means the self-driving car can steer clear of wandering pedestrians, avoid cars that run red lights, and even understand cyclists’ signals.

Watch Logic: Bayesian Robots to find out more about how robots use Bayes’ theorem to learn about the world around them.