Lights, Camera,
Vehicle Safety


Srinivasa Narasimhan, an associate professor at the Robotics Institute at Carnegie Mellon University, is part of a team working on programmable automotive headlights to make poor visibility conditions on the road easier to manage by sensing, reacting to, and adapting quickly to any environment while moving at highway speeds. “We’re looking at post-processing images,” he says. “If the rain or snow is so bad that you can’t instantly post-process it, you want to control the lighting to improve the situation. That’s where the headlights came about.”

Initially, they had two cameras and a small projector but it was very slow. “We tried to show some simple proof of concept in the lab and realized the big challenge is keeping it working while moving at high speed,” Narasimhan says. “To do this, we had to bring down the latency of the system. Right now, the reaction time is 1-1.5 milliseconds, which is 500 times faster than an average driver’s reaction time if we want to stop at a red light or stop sign.”

The programmable headlight is a system consisting of a camera, processor, and spatial light modulator. Image: Carnegie Mellon University

One big change from the initial version is that the camera and the light source are now co-located. “Co-location means essentially at the same spot so we did that by using a beam splitter and it allows us not to have to calculate the distance of objects in front of the car,” he says. “Some other methods can be pretty erroneous and using erroneous data to process and change the lighting would make it flicker much worse than an ordinary headlight. We also wanted to make sure that the data flows well from the camera. You don’t want any buffering or road hums anywhere, just moving the data from one component to another. Optimizing that speed was very important and it’s now 10 times faster than a year and a half ago.”

One amazing part of this story is that they’re still using basic technology. The camera he categorizes as “a regular Sony,” the processor is an Intel, much like you’d have in any PC, and the projector is essentially a DLP that they “hacked up.”

A View from the Hood

They’ve already shown actual performance on a vehicle. “We put it on the hood and drove it around,” he says. The anti-glare improves visibility of the road and the system highlights lanes where the road doesn’t have markings in it and can even spot bicyclists, he says. “The capability of the system allows us to do anti-glare at very high speed—70 mph even and it can deal with pothole turbulence.”

Still, challenges coming up include making the system fit in the headlight slot. “Right now our sizes are for the space of a Ford F-150 pickup truck and we obviously want to miniaturize it,” he says.

For Narasimhan, this multi-year odyssey has been incredibly exciting, especially when he considers the possibilities. “To its fullest potential, it could improve the stress of driving at night and could increase beam brightness,” he says. “Many people need to be on the road and we want to give them more confidence and better reaction time. That means we need to keep on working on improvements.”

Eric Butterman is an independent writer.

Many people need to be on the road and we want to give them more confidence and better reaction time.

Prof. Srinivasa Narasimhan, Carnegie Mellon University


December 2014

by Eric Butterman,