Smart headlights allow drivers to use high beams safely

Srinivasa Narasimhan is an associate professor of robotics. (credit: Srinivasa Narasimhan) Srinivasa Narasimhan is an associate professor of robotics. (credit: Srinivasa Narasimhan) Robert Tamburo is a project scientist in the Robotics Institute. (credit: Robert Tamburo) Robert Tamburo is a project scientist in the Robotics Institute. (credit: Robert Tamburo)

Nearly everyone has been told at some point in their life to drive safely. In today’s society, this reminder is often a warning about speeding or texting while driving. Despite the importance of these dangers, though, many drivers are unaware of one of the biggest driving hazards — the dark. Robert Tamburo, a project scientist in the Carnegie Mellon Robotics Institute, along with Srinivasa Narasimhan, an associate professor of robotics at Carnegie Mellon University, are working to reduce this hazard by developing a smart headlight that allows drivers to use high beam headlights without blinding oncoming drivers.

Despite a severe decrease in traffic, over 40 percent of all fatal car accidents occur at night. According to the National Organizations for Youth Safety, driving at night is a top cause of car accidents due to the inability of the eye to respond properly to switches between bright lights and darkness. The combination of bright headlights from oncoming cars and surrounding darkness can cause drivers’ eyes to make continual adjustments that impair vision.

Tamburo and Narasimhan have developed a smart headlight that reduces the risk behind high beam headlights. “With our programmable system.... we can actually make headlights that are even brighter than today’s without causing distractions for other drivers on the road,” Narasimhan said in a university press release. The smart headlight is comprised of a Digital Light Processing (DLP) projector, as opposed to standard headlights, which are comprised of standard light bulbs or clusters of LEDs. The DLP projector splits the light into a million minuscule beams that can each be controlled independently by a computer.

This increased control allows the headlight to recognize, track, and respond to oncoming cars as well as road signs and drops of precipitation. After sensing an object, the headlight can turn off small sections of the high beam to prevent glare; this includes blacking out the light that would shine into the eyes of oncoming drivers, as well as the light that would reflect off precipitation back into the drivers’ eyes. In addition, the headlight can increase brightness when beneficial; for example, upon the perception of street signs, traffic lane lines, or moving objects on the side of the road.

Even with these changes in headlight brightness, the headlight intensity appears virtually unchanged to the driver due to the small size of the adjustment. The system latency — the time it takes for the projector to adjust illumination — is also small, falling between one and 2.5 milliseconds. By employing such quick alterations the headlight minimizes visual distraction for the driver from variations in light intensity.

The smart headlight offers a new way to minimize the risk of nighttime driving, yet there is still work to be done. The current model of the headlight is larger than standard headlights, which limits its application. At its current size, the headlight can only be used for larger vehicles such as trucks and buses, which are prone to glare. Future miniaturization of the headlight could lead to more widespread usage.

Along with Tamburo and Narasimhan, the research team includes Takeo Kanade, professor of computer science and robotics, Anthony Rowe, assistant research professor of electrical and computer engineering,master’s students Abhishek Chugh, Subhagato Dutta, and Vinay Palakkode, and Eriko Nurvitadhi and Mei Chen of Intel Research.

This research is part of the Technologies for Safe and Efficient Transportation Center, a U.S. Department of Transportation University Transportation Center at Carnegie Mellon, and was backed by Ford Motor Company, the Intel Science and Technology Center for Embedded Computing, the Office of Naval Research, and the National Science Foundation. Tamburo, the lead engineer, presented the research on September 10 at the European Conference on Computer Vision in Zurich, Switzerland.