PROVIDENCE, R.I. [Brown University] — The eighth floor of Brown’s Sciences Library, home of the Humanity Centered Robotics Initiative labs, is buzzing this summer.
There’s the figurative buzz of students working feverishly on summer research projects. But there’s also a literal buzz, generated by small quadcopter drones conspicuously hovering in the hallways.
The drones belong to a team of undergraduates who have spent the summer working to refine an introductory robotics course taught by Stefanie Tellex, an assistant professor of computer science. The class teaches students to assemble their own quadcopter drones from kits funded by Amazon, and then program them to fly on their own. The idea, said Tellex, is to use the drone as a fun and inexpensive platform for teaching basic concepts in robotics — from the complexities of integrating hardware and software to the potential problems that arise when programming autonomous machines.
Tellex tasked the student team — rising sophomores Luke Eller, Theo Guerin, Garrett Warren and Sophie Yang — with improving the drones and enhancing the concepts students will learn when they take the class in the fall. The team worked under the auspices of a Karen T. Romer Undergraduate Teaching and Research Award (UTRA), which provides students with funding to stay on campus and work on research projects over the summer.
“Last year was the first time that course ran, and we inherited that stack of software and the drone design used last year,” Eller said. “Our job is to improve that software, extend the capabilities of the drone and rework the course based on feedback from the students who took the class last year.”
The drones used in the course are small and fairly rudimentary, which Tellex says is by design. She wanted a platform that would be safe and inexpensive, but still sophisticated enough to teach key concepts. Tellex and her students dubbed the machines “PiDrones,” because the “brain” of each one is a Raspberry Pi, a small but mighty computer processor available for around $35. The processor is capable of running ROS (Robot Operating System), the go-to system for driving research robots.
For their summer project, the UTRA team worked to refine algorithms that are critical in making the drone fly autonomously — without the need of a human operator joy-sticking its every move. For example, in order for the drone to move by itself from one place to another, it has to have some sense of where it is at any given time, and where it’s going. That requires an algorithm called SLAM, which stands for “simultaneous localization and mapping.” The students helped to refine the computer code that enables the drones to perform that critical task.
Another key algorithm helps the drone’s computer brain interpret the data supplied by its sensors. Each drone has a downward-facing camera, range sensor and an inertial measurement unit that help it to assess its surroundings. But those sensors aren’t perfect. They supply noisy data with substantial margins of error. The drone’s central processor needs to be able to deal with that error, and choose and appropriate course of action. That’s done through an algorithm called a Kalman filter, which creates probability distributions that make noisy sensor data actionable.
Developing and refining these algorithms won’t just allow future students to create autonomously-flying drones, said Warren, one of the UTRA team members. It will also provide students with broadly applicable tools to develop all kinds of robots that could operate in the real world.
“We’re leaving stencils of what we’ve done to let the students who take the class learn it for themselves,” Warren said.
A summer well spent
All of the team members said that the process of thinking about how to teach robotics to others has helped them to learn robotics for themselves.
“For me this has been important because this is my first time focusing on advanced topics in robotics, not just tinkering around,” Guerin said. “We’re getting into more advanced mathematics and seeing how those mathematics apply [to robotics].”
Guerin added that it was gratifying to learn that the concepts they’re working with go well beyond the drone. The Kalman filter, for example, was used to help the Apollo astronauts land successfully on the Moon. “That’s the beautiful thing,” he said. “When you step back and see that these concepts apply to a lot of things.”
Yang shared similar sentiments about her work with the team, and said the experience revealed new academic and career possibilities.
“Coming into this project, I didn’t know much about robotics,” she said. “But after exploring a little bit more, I’ve learned more about computer vision too, and that’s something that’s really interesting to me. So doing these research projects really opens up a lot of different pathways for exploring what you might be interested in studying.”
Warren said he appreciated the summer research program for enabling a deep, dedicated dive into a single subject.
“When you have four classes, you really can’t go all in [on one topic],” he said. “But when you have eight hours a day to focus on one thing, it feels like a lot can get done. It’s a set time to delve into and understand things, and be motivated by whatever interests us to pursue further.”
All four students will serve as teaching assistants when the class is offered in the fall, so they’ll have a chance to see all of their work in developing the course come to fruition. And it might not end there. Tellex hopes to one day offer the class online and mail drone kits to students around the world.
In that way, these four students may wind up helping untold numbers of future students to get a foothold in robotics. Not a bad way to spend a summer.