Self-driving cars are very good at working out what’s around and in front of them, but they’re not so adept at determining whether another vehicle is coming around the corner. Drivers might notice distant headlights, but autonomous cars can’t.
But researchers at MIT’s Computer Science and Artificial Intelligence Laboratory may have found a solution. They’ve developed a new algorithm that, in addition to detecting hidden objects and people, determines their speed and trajectory in real-time.
It’s called CornerCameras, and it’s capable of detecting the fuzzy shadows that appear in lines of sight from objects around a corner. By doing so, the system effectively turns edges into “cameras that reveal the hidden scenes beyond them”.
Unlike other systems, CornerCameras doesn’t involve shining light onto empty space. It also works in challenging weather conditions and, because it’s smartphone compatible, works with a range of off-the-shelf products.
Researchers will now address the system’s challenges – namely low-light (or rapidly changing) conditions and the fact that, with smartphone cameras, the signal gets weaker as you get further away from a corner.