Derq's artificial intelligence is designed to do two things. It makes accurate inferences about vehicles that are going to run red lights before they do it, and it broadcasts immediate warnings to vehicles in the area. It also predicts the intentions of pedestrians and other vulnerable road users and alerts vehicles if a conflict is anticipated.
The system must make such forecasts with enough time to give motorists a two-second warning.
"The metric we're seeing and shooting for is an ability to predict true positives 90-plus percent of the time, and we're doing that with extremely low false positives," said Will Foss, director of business development and partnerships for Derq, which received an $80,000 grant from PlanetM.
Foss said the company plans to expand testing in Michigan and elsewhere in the U.S. next year. Though the company remains agnostic on whether its V2X, or vehicle-to-everything, warnings are sent via Digital Short Range Communications or over cellular networks, Derq has demonstrated its software while using Qualcomm's cellular chipsets, which can alert autos that have V2X built in or have an aftermarket receiver.
An unforeseen benefit has developed since Derq started testing. As the use of e-scooters has sprouted around the city, the company's data has given city transportation officials and safety advocates insight into where and how they move around.
"Before, it was bikes and pedestrians," Foss said. "Now there's a whole new class of vulnerable road users that are kind of on the sidewalk, sometimes not, and kind of like a bike, but not a bike. So we're looking at their usage and behavior, and learning to predict and understand how they're impacting risk."