When Mihir Garimella thinks of fruit flies, he thinks of escape artists.
His unusual interpretation of household pests has served him well. The incoming freshman at Fox Chapel Area High School built a flying robot inspired by the fruit fly‘s capacity to escape moving objects, which was selected last month by judges of the international Google Science Fair as one of 30 finalists in the Americas. That’s no mean feat when you are competing with thousands of science-heads from the most northern reaches of Canada to the tip of Tierra del Fuego.
Mihir, the only finalist from Western Pennsylvania, will learn Aug. 6 whether he will be one of 15 participants selected by Google to fly out to its headquarters in Mountain View, Calif. These will be selected from a total of 90 finalists from around the world.
In a crisp, blue button-down and pressed chinos, he looks like a Silicon Valley entrepreneur. He sounds like one, too. Since age 7 or 8 he has been breaking up his ideas into bite-sized commands that are computer-codable, and he talks in the same way. He even offers to translate his computerese into layman‘s English.
But, as a 14-year-old living with his family in O‘Hara, he is not allowed to stay up all night programming. He is not even allowed to bring his laptop up into his room: His mother prefers to know what’s happening on her kids‘ computer screens. When he is working on projects, Mihir’s laptop is usually filled with the visual chatter of code.
Often, he finds corners -- either in his own daily life or that of others -- and then uses programming to get out of them. His uncle, a pathologist, complained of the tedium and imprecision of analyzing images of brain tumors, so Mihir automated the process. He was in seventh grade. When he was 11, he’d created Robo-Mozart, which tuned his violin without the risk of breaking a string. At 13, when speaking to his grandparents through FaceTime, he longed for the smells of jasmine and spices that he remembered from their house in Bangalore, India, so he made ScentIt, a machine that embeds smells in video clips. He has met President Barack Obama (for winning a technology award for Scentlt) and coauthored a paper in the journal Brain Pathology. There is even a minor planet near Neptune that carries his name.
He explains that his FlyBot emerged from a mistake. Last summer, when his family members visited his grandparents in India, they left bananas on their kitchen counter. For fruit flies, the rotting fruit was both a feast and a nursery, and by the time the Garimellas returned, the kitchen was clouded with swirling insects. But when Mihir tried to squash them, they easily avoided his hand.
Watching them, he realized that the fruit flies were succeeding where flying robots had failed. In 2011, after an earthquake in Japan, a small helicopter-like robot called a quadrotor was sent into a building at Tohoku University to map the damage. It had a stellar record of avoiding stationary obstacles as it flew, but it could not deal with moving objects. “They had to modify the environment before they sent it in,” Mihir said. “They removed things that could fall or collapse and pose a threat to the robot.”
The video footage used to perceive moving objects, he said, took too long to read, making it impossible for the robot to react fast enough to alter its flight path.
Yet here were these fruit flies avoiding his murderous snatches and swats. Their brains are one-millionth as complex as a human’s, and Mihir wondered whether simplicity is their secret. So he used two simple infrared distance sensors, attaching them to a small commercial quadrotor, and programmed the robot to buzz in the opposite direction of an oncoming threat.
“The comparison to a fly is very figurative,” says Sanjiv Singh, a professor at Carnegie Mellon University‘s Robotics Institute, in whose lab Mihir is working this summer. All eyes are passive sensors, explained Mr. Singh, while the FlyBot tracks movement by putting out infrared beams.
Mr. Singh has led Mihir away from the fruit fly hypothesis, enlisting him instead to figure out how a robot can zoom around the lab without hitting anything.
“It’s cluttered with all kinds of objects. Some are thin, some are thick, some are textured, some are shiny. They are not easy to recognize,” said Mr. Singh. “It is a difficult problem, and he is coming up with creative solutions.”
Although Mr. Singh does not see the FlyBot as the best tack for Mihir to take, the robot is up for the Google Science Fair Computer Science Award, which could win its creator $25,000 and a year of mentoring to pursue his project. Mihir also stands to win the grand prize, which includes a trip with National Geographic to the Galapagos Islands and a $50,000 scholarship.
To read more about the Google Science Fair, go to googlesciencefair.com.
Eric Boodman: firstname.lastname@example.org or 412-263-3772. Or on Twitter, @EricBoodman