Cruisin' In Duckietown: MIT's Rubber Duckie Taxis Can Self-drive
For our rubber duckie friends, cruisin' along Duckietown has never been slicker than this.
Massachusetts Institute of Technology (MIT) has opened a nifty robotics class this spring to teach students about self-driving cars without the perils of running around with actual 3-ton vehicles.
The result? Students produced fifty duckie-adorned taxis that can self-drive and navigate the roads of Duckietown -- a model city located in the sovereign state of Duckieland.
Duckiebots In Duckietown
Created at the university's Computer Science and Artificial Intelligence Laboratory (CSAIL), the rubber duckie taxis have individual on-board cameras and can even recognize traffic signals as well as road markings. This is quite similar to what auto and traffic engineers face in the real world, according to Pop Science.
What makes the Duckiebots stand out? Typical automated and semi-automated cars currently in development rely on multiple cameras and/or preprogrammed maps to determine which route to take Duckiebots, however, can navigate with only one camera and requires no preprogrammed map.
In order to create a consistent and accurate system, students in the course were put to test by making trade-offs.
For instance, between cheaper hardware and sophisticated algorithm or more reliable hardware and simpler algorithms, which is the right decision?
Andrea Censi, one of the lead scientists of the Duckietown project, said they thought about key issues such as co-design and integration.
They wanted to tackle questions such as the following: how do we ensure that systems that grew separately could work side-by-side? How do we create systems that make the most of performance while using the same materials?
"It's a delicate balancing act in weighing the relative importance of different infrastructure elements," said Censi. It's a great way to challenge students to think hard about their design.
Duckietown's Mission And Vision
Duckietown project researchers aim to work with roboticists all over the world to integrate the course's open-source materials and $100 vehicle design into other schools. In other words, create more Duckietowns in other places.
Liam Paull, co-leader of the Duckietown course together with Censi, says they believe that a tool such as this will create a common language and platform for experts to build on.
He hopes the project will bridge the gap between computer scientists and bring them together to turn autonomous vehicles into reality.
Watch Duckiebots cruise around Duckietown below.