Duckietown started out as an engineering course at MIT that was designed to teach students about the world of machine learning and autonomous vehicles. The platform was modular and scalable, and used to transport small rubber duckies around a model town using a single onboard camera and no pre-programmed maps.
Two years later, MIT engineers have started the Duckietown Foundation, which aims to bring the platform into classrooms and homes in an effort to provide an introduction into the world of robotics and AI.
Touted as a “playful learning experience,” Duckietown is already being used and has taught over 700 students in 10 different countries, including MIT, ZTE (Zurich), TTI Chicago, the University of Montreal, and NCTU, among a host of others. The learning experience involves students building and programming their own Duckiebots to navigate around a scalable city, creating new functionalities and revamping existing ones.
The Duckiebots themselves are outfitted with only a single HD camera, which it uses to garner information on where to travel in the simulated town by detecting lanes and their orientation in those lanes. Each autonomous Duckiebot features an Raspberry Pi 3 to process the visual data and then engages the robotic vehicle on where to move using a DC motor shield and a pair of DC motors.
The Duckietown platform can be configured using a modular approach that allows Duckiebots to perform different functions. Those functions are designed to translate AprilTags placed around the town to denote topography types, traffic signs, and lights. These visual cues help the autonomous vehicle to navigate through the village to its destination.
Duckietown is currently being crowdfunded on Kickstarter in kit form, with the starter version running $399, which gets you a single Duckiebot and enough tiles to form a loop track. There are also options for donating a kit to kids in need and education packages for teachers and schools for use as learning materials.