Amazon Web Services Inc., a subsidiary of Amazon.com Inc., today announced it’s open-sourcing the device driver software at the heart of its AWS DeepRacer robots, its 1/18^th-scale autonomous racing vehicles designed to demonstrate machine learning technology.

AWS DeepRacer provides a fast way to learn about machine learning programming by putting it into practical terms using autonomous vehicles as a real-world example of its use.

In DeepRacer’s case, it is an Ubuntu Linux-based computer on wheels powered by the Robot Operating System. It can be bought on Amazon.com as a kit starting at $319. With today’s announcement, Amazon is offering a 25% discount on AWS DeepRacer ($100 off) and AWS DeepRacer Evo ($150 off) until May 27. Either is required to get started.

There is also an AWS DeepRacer League where developers write their own machine learning “brains” to operate in the robots and compete against each other on real and virtual cloud-based simulated tracks – given the pandemic – for glory and prizes.

AWS DeepRacer uses a type of machine learning called reinforcement learning, an advanced learning model that learns complex behaviors without requiring labeled training data by ingesting sensor data. it can make short-term decisions while optimizing for longer-term goals.

Amazon provides the ability to train models against a virtual track that matches a real-world track to build up the machine learning scripts for the robot. That way training, testing and iteration can happen quickly before the rubber meets the road – though eventually it’s more fun to put wheels on cement.

By open-sourcing AWS DeepRacer and providing a GitHub repository, developers are no longer locked into a racetrack-bound DeepRacer. They can override the default behavior and teach it to do whatever they want and even employ a large number of sensors and other devices.

Want the car to drive to a particular point in a grid, connect to a trailer and deliver it to another location, such as GrubHub in your home? Do you want it to follow your cat around your house? How about turning the AWS DeepRacer into one of those annoying alarm clocks that run away from you in the morning so you can’t hit the snooze button?

AWS is offering six example projects to get developers started including Follow the Leader, Mapping, Off Road, RoboCat, DeepBlaster and DeepDriver. RoboCat is particularly amusing since it uses the DeepRacer and an infrared camera in an attempt to frighten mice in the house by driving at any movement seen at night. (No mice were actually frightened during the demo.)

Machine learning has myriad applications beyond robot racing cars that extend into autonomous cars on roadways, predicting the weather, assisting with package delivery, pharmaceutical research, financial technology and many more industries.

Photo: Amazon