High-beam headlights cut through thick precipitation like a knife cuts through butter, but they’re a literal eyesore to oncoming drivers. That’s because most commercially available systems don’t automatically adjust to compensate for daytime brightness, and because they have a habit of producing glare.

The problem spurred a team at Nvidia’s Drive Labs division to investigate an AI solution, which they detailed in a blog post and accompanying video this afternoon. They say that their camera-based system — AutoHighBeamNet, a component of the AutoDrivingBeam module in Nvidia’s Drive Software — can automatically generate control outputs for cars’ high beams using signals derived from road conditions.

AutoHighBeamNet ingests image data from a front-facing camera, which it processes and outputs to AutoDrivingBeam in a format that can be customized by carmakers to account for speed, environmental conditions, and more. Rather than generating control signals based on lux levels of other light sources, Nvidia says that AutoHighBeamNet reacts to active vehicles in the perceived camera frame. (Roadside parked vehicles with their lights turned off are ignored.)

Nvidia Drive AI

The control signal can take on one of two modes: auto high beam (AHB) mode, which provides binary on or off control, and adaptive driving beam (ADB) mode, which provides control of individual high beam LED arrays to create “glare-free” zones. In AHB mode, the vehicle’s high beam lights automatically turn on in poorly illuminated nighttime driving conditions, but they turn off and switch to a low beam when an active vehicle is detected. In ADB mode, when another car is detected, the high beam is shaped in a way that prevents glare to active vehicles by dimming individual LEDs in the high beam LED array headlamp.

Nvidia says an API will provide AHB mode support in version 10.0 of Nvidia Drive, alongside information carmakers need to define their own policies for auto high beam control.

The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here