July 26, 2012 | 8
Drive through a pounding thunderstorm or heavy snowfall at night and you’ll notice that your headlights illuminate the maelstrom of raindrops or flakes more than they shed light on the road ahead. This safety hazard can be greatly reduced by anticipating the movement and velocity of the drops or flakes and shining the headlights into the spaces between them, according to a team of researchers who’ve built just such a “smart” headlight system. (pdf)
The headlight is actually an array of bulbs. The system uses a digital camera to track the motion of individual raindrops or snowflakes and then applies a computer algorithm to predict where each bit of precipitation will be a few milliseconds later. It deactivates bulbs that would otherwise illuminate the drops or flakes in their predicted positions.
“We have introduced a headlight system that can see through rain” and other precipitation, says Srinivasa Narasimhan, an associate professor at Carnegie Mellon University’s Robotics Institute. Narasimhan’s team worked with researchers from Texas Instruments and France’s Paris Institute of Technology, with funding from the Office of Naval Research, the National Science Foundation, the Samsung Advanced Institute of Technology and Intel Corp.
Narasimhan is the first to admit that the concept sounds a bit farfetched. The key is that even a sheet of heavy rain is still mostly empty space. “When we see rain, we see streaks, so it looks like the precipitation is inhabiting the air very densely,” Narasimhan says. “But if you look at an instantaneous snapshot of rain it’s actually very sparse. This is the case even in a thunderstorm.”
He and his colleagues have built what he refers to as a prototype “reactive illumination” headlight. Its camera captures an image every 8 milliseconds and adjusts the bulbs in the headlamp accordingly within 13 milliseconds. Narasimhan claims it reduces the visibility of rain four meters away from the light source by about 70 percent when a car is moving at 30 kilometers per hour. The prototype reduces the visibility of snowflakes, which drop more slowly and tend to be larger than raindrops, by 60 percent.
Of course, for a smart headlight to be useful, it would have to be accurate while traveling at high speeds. Narasimhan wants to test his headlights in cars traveling at least 95 kilometers per hour. “From simulations the we’ve done we know the system as a whole needs to operate at least four times faster than it currently does in order to provide drivers with a 70 percent visibility enhancement in a thunderstorm,” he says. “As you’re driving really fast the rain is coming at you faster, and therefore your system has to be much faster.”
The researchers hope to cut system latency by developing faster ways of transferring images from the camera to the microprocessor running their algorithm and from that microprocessor to the headlight. This need for speed might be accompanied by a headlight using arrays of light-emitting diodes (LEDs) rather than bulbs. These LEDs might even be combined with image sensors on a single chip to facilitate high-speed operation, according to Narasimhan.
Another issue is the light produced by other vehicles on the road. If you are driving on a road and oncoming traffic is likewise using reactive illumination headlights, those lights would be trained on the spaces between the precipitation and not increase the visibility of those raindrops or flakes for you, Narasimhan insists. If, however, you are the only one using such headlights, your ability to see through the precipitation would be mitigated by other cars indiscriminately lighting up everything around them.
Either way, Narasimhan says it’s likely going to be another two or three years before his smart headlight system is road ready.
Images courtesy of Carnegie Mellon University