Microsoft unveiled the Kinect two years ago as an add-on for the Xbox. The infrared motion-detecting sensor was designed to enhance video games, to turn you, as Xbox puts it, into your own controller. But immediately, of course, researchers at MIT started hacking the things.
A tool initially intended for your virtual dance moves on Michael Jackson: The Experience had obvious potential for scientists as well to study how people move through all kinds of scenarios and spaces. Researchers at MIT’s Senseable City Lab figured the simple gadget–available for about $80 at the scientific warehouse of Best Buy–could change how scientists study all kinds of crowd behaviors in our increasingly congested cities.
They bought some Kinects and mounted them on campus. Now their project–-Kinect Kinetics, they call it–-has demonstrated a new technique for tracking human behavior that could ultimately help us build smarter airports, train stations, public plazas, shopping malls and supermarkets. With 60% of the world’s population projected to cram into urban areas by 2030, this is no trivial invention.
“When you have this data, you can really use it to better design spaces for human interaction,” says Carlo Ratti, director of the Senseable City Lab. “This is the most exciting thing.”
The Senseable City researchers aren’t simply modeling virtual pedestrians (although there have been some impressive advances on that front, too). They’ve been using the Kinect to track actual people moving through real spaces. And that kind of information can help populate more accurate computer crowd models.
Until now, researchers have studied this kind of human movement in a remarkably low-tech way: by manually annotating, sometimes frame by frame, video footage of people moving through space. This sounds about like the worst job in all of science. “It is,” laughs Stefan Seer, one of the researchers on the project, “for students.”
Automatic video tracking can also now digitally extract trajectories of human movement from video footage. “But the state of the art is still in this phase where it will not work with large crowds, at high densities,” Seer says.
At MIT, he and several colleagues mounted three Kinects from the ceiling of the school’s “Infinite Corridor,” an infamously long 825-foot hallway (it has its own Wikipedia page!) that connects main buildings across the campus. It’s a kind of pedestrian superhighway, making it a great place to test sensors for tracking the movement of tons of people. Seer and his colleagues mounted the three sensors close to each other, so that each one would pick up the passing pedestrians below as they moved down the corridor and out of range of the previous sensor. Their algorithms were then able to stitch together data from all three sensors continuously following any given 3-D avatar below.
Some of what the researchers learned was obvious: Pedestrians in the Infinite Corridor, for instance, clearly self-sort into two passing lanes, with people traveling on the right side of traffic just as they do in cars (random trivia: pedestrians in countries where they drive on the other side of the road, like England, do the opposite). But to really program finely tuned computer models of human behavior, researchers need more data than you can collect with your eyes. How do people move under different circumstances, or with different wayfinding cues? How do they travel in relation to each other? And what happens when you drop an obstacle in their midst?
The MIT researchers tried all of these scenarios, in some cases having colleagues block traffic in the middle of the hallway to observe how people would move around them (as it turns out, pedestrians move like atoms, magnetically repulsed away from obstacles and each other). A long corridor is also a pretty simply setting. Researchers could learn a lot more mounting these things in chaotic train terminals, or at the airport around Thanksgiving. The findings could then inform both the structural details of how we construct such places (where to put the escalator) as well as their more subtle design touches (like the signage leading you to baggage claim).
“In the end,” Ratti says, “the goal is to design places where you can better orient yourself and where you can have a more sociable experience.” He’d even like to put these sensors inside private homes, maybe learn something about how you move around your kitchen, how you use your home appliances. Looked at that way, this idea has applications not just for large crowds in public places, but for elderly individuals who simply want to age in their own homes.