Don't let this fleeting sunshine lull you into a false sense of security. It's February in New England, and yes, there will be more snow to come. So take a moment to appreciate one wonder this winter brings: an autonomous robot that can catapult snowballs at your friends, enemies or anyone else you see fit.
Frost is a robot built by students from Olin College. And using computer vision, sensor fusion and artificial intelligence, Frost can detect people and automatically launch snowballs at them with accuracy equivalent to or better than that of a human.
Frost was developed in 8 weeks, as part of the Principles of Engineering course during the fall semester. The students behind the snowball bot are Cedric Kim, Jeremy Garcia, Daniel Daugherty, Kevin Zhang and Kevin Guo. Over the course of four sprints, using about $775, the group was able to take Frost from proof of concept to the sizable prototype they have today.
"The project itself was that we had to integrate mechanical, software and electrical engineering components, and everything had to be operated by a micro-controller," Garcia said. "So it was a pretty broad prompt for us."
The automated catapult is outfitted with LIDAR laser sensors and Xbox Kinect's body tracking technology, which enables Frost to hurl spheres of snow up to 40 feet at a target with a 6-inch margin of error. All you have to do is arm it with the snowball ammo needed.
Zhang told me, "If you watch the unedited video with about 20 demos... it hits him in every single shot, with the exception of one shot, which hit him in the right arm instead of the center of the chest. That’s where the 6-inch error margin comes in. From all the testing, including the ones not filmed, it was at 98 percent or so in terms of accuracy."
For now, the students don't intend on commercializing Frost. However, they're thinking about working on a compactor component to the robot, which they initially nixed from their prototype, so Frost can eventually make and launch snowballs on its own.
