At last month’s Computer Vision and Pattern Recognition Conference in Portland, OR, Marcus Brubaker of Chicago’s Toyota Technological Institute (TTIC) demonstrated a GPS-free geolocation system for cars. It works through the use of two simple cameras, which collect video data that is then compared to an OpenStreetMap area map. Through a process of elimination that takes place within an average of 20 seconds of driving, the system is able to figure out exactly where the car is. During tests conducted in Germany, it located cars to within 3 meters of their actual positions. Brubaker says that even in a dense metropolis, such as Manhattan, the system is sensitive enough to distinguish between city block sizes.
Subscribe for counterintuitive, surprising, and impactful stories delivered to your inbox every Thursday
What’s the Big Idea?
While GPS has helped to revolutionize transportation, it isn’t perfect by any means: It can “lose” a car when it passes through a long tunnel, for example. Accuracy will become even more important once driverless cars finally arrive on the roads, says Carnegie Mellon researcher Junsung Kim. He notes that the TTIC system may require more work to ensure that cars know which lane they’re driving in.