On the off chance you haven’t seen it yet, the UPenn GRASP Lab has just released yet another impressive video of their performance quadcopters. This time they’ve got a new “nano” version that’s smaller, lighter, and capable of flying in formation. It’s like synchronized swimming, but with more buzz:
Now as usual, the internet comments on this latest quadrotor development have largely consisted of “WOW!” and “Good heavens, they’ll kill us all!” But as for me, once I’d retracted my dropped jaw, I started trying to figure out how it all worked. Unfortunately, I haven’t found any real documentation on this project beyond the videos posted by PhD candidate Daniel Mellinger. I just might have to send him an email…
But never fear, I have at least discovered how they’re tracking the quadcopters. Did you notice those camera-like devices mounted along the walls? And those funky red ring-lights surrounding their lenses? Well those are VICON motion capture cameras:
These devices by themselves are quite impressive. Much like traditional video cameras, each of these units contains a sensor with a certain number of megapixels. However, VICON sensors are designed for fast frame rates (up to 2000fps), high resolution (up to 16MP, more than seven times the resolution of 1080p HD Video), and sensitivity to the red/infrared light emitted by their ring light strobes. Why red light? Well, these aren’t your typical video cameras. Their purpose isn’t to capture a full-color image, it’s to capture points of light coming from passive reflectors. In fact, what each camera sees looks like a star map of sorts. Once multiple cameras are setup and calibrated, sophisticated software can measure and track the position of each reflector in real time.
So in the video above, it appears that each quadcopter has at least two reflectors attached to its top surface. The perimeter cameras can then measure the position and orientation of each unit and relay that information to some kind of controlling computer. What I still don’t know is how the software distinguishes between each quadrotor. Perhaps VICON has reflectors which can be distinguised by the precise wavelength of light at which they reflect? Or maybe the computer is just smart enough to know that the same set of points represents a certain unit from one frame to the next?
I’m also wondering how each quadcopter is controlled. Some type of ZigBee wireless link perhaps? And does the main computer handle everything? I suspect there must be a certain amount of control embedded in each unit. Perhaps they’ve all got accelerometers and MEMS gyros keeping them straight and level. Still a lot of unanswered questions. But in the meantime, enjoy this video on a radically different use of the VICON technology: