Self-Driving Cars Meet Rubber Duckies in ‘Duckietown’
"Duckietown" is model town created by an MIT Computer Science and Artificial Intelligence Laboratory class where students built a fleet of 50 "Duckiebot" self-driving taxis that can drive around the model city with just a single on-board camera and no pre-programmed maps.
If my two-year-old daughter was a roboticist, this video would be straight out of her dreams: miniature self-driving taxis taking rubber ducks around “Duckietown.”
But “Duckietown” is actually a model town created by an MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) class where students built a fleet of 50 “Duckiebot” self-driving taxis that can drive around with just a single on-board camera and no pre-programmed maps.
The students created algorithms to read traffic signs and notice pedestrian-ducks, and learned to integrate different disciplines like control theory, machine learning, and computer vision into their systems. And, as always, their were trade-offs that had to be made. For example, is it better to have sophisticated algorithms with cheaper hardware, or simpler algorithms with more reliable hardware?
“We thought about key problems like integration and co-design,” says Andrea Censi, a research scientist from the Laboratory for Information and Decision Systems (LIDS) who co-leads the course with CSAIL postdoc Liam Paull. “How do we make sure that systems that developed separately will work together? How do we design systems that maximize performance while sharing resources? It’s a delicate balancing act in weighing the relative importance of different infrastructure elements.”
But the coolest part of “Duckietown” is what the creators want to do with it. They hope to work with roboticists around the world to incorporate their open-source teaching materials and $100 Duckiebot design into schools’ programs.
“We believe a tool like this will help create a common platform and language for researchers to build on,” said Paull. “We hope this will make it easier for computer scientists to continue to work together to bring autonomous vehicles into the real world.”
Of course, this isn’t the first time we’ve seen mini self-driving cars from MIT. In 2015, students in the Rapid Autonomous Complex-Environment Competing Ackermann-steering Robot (RACECAR) course were tasked with building the fastest self-driving car possible on a 1:10 scale. Teams then raced their cars around a 515-foot race course for bragging rights, with three of the four teams successfully completing the race.
Surely “Duckietown” will be instrumental to inspiring future generations of roboticists.
Here are a couple more videos of Duckietown, which you can see in person on Saturday at MIT’s Open House from from 10 a.m. to 3 p.m.