Amazon Open Sources DeepRacer Device Software

AWS DeepRacer Software Now Open Source

Amazon initially launched AWS DeepRacer in 2018 as an engaging platform for developers to explore the principles of machine learning. The project has since expanded to include competitive DeepRacer events. Now, Amazon has announced a significant update: the open sourcing of the software powering these miniature vehicles.

Core Technology and Open Source Foundations

The DeepRacer car fundamentally operates as a compact computer, utilizing both Ubuntu Linux and the Robot Operating System (ROS) – both of which are already open-source technologies. Amazon anticipates that releasing the device software to developers will foster innovation. This will allow for modifications to the car’s standard operational parameters.

According to a recent blog post, the company stated, “By open sourcing the AWS DeepRacer device code, you gain the ability to swiftly and easily alter the default behavior of your race car, which is currently focused on track navigation. Do you envision implementing defensive maneuvers to prevent other cars from passing? Or perhaps deploying a bespoke algorithm to enhance the car’s speed between two points? All that is required is imagination and coding expertise.”

Expanding the DeepRacer Ecosystem

Since the initial introduction of the cars in 2018, Amazon has established both in-person DeepRacer leagues and virtual racing competitions. The league structure was recently revised last month to broaden participation and encourage wider adoption of the technology. The addition of an open-source element is expected to further stimulate interest.

Developers now have the opportunity to personalize the cars, unlocking new functionalities and applications previously unattainable.

Educational Goals and Reinforcement Learning

The overarching objective of this initiative is to educate developers on the fundamentals of machine learning. As Marcia Villalba of AWS explained in a blog post last month:

“AWS DeepRacer is a 1/18th scale autonomous race car designed to evaluate reinforcement learning models through virtual racing in the AWS DeepRacer console or physical races at AWS events and customer locations. It’s accessible to developers of all skill levels, even those without prior ML experience. Learning RL with AWS DeepRacer provides a fun and competitive environment to gain practical machine learning experience.”

Getting Started with the Open Source Project

For those interested in customizing the car’s software, comprehensive documentation is available on GitHub and the AWS DeepRacer Open Source page. Six sample projects are provided to facilitate initial exploration and development.