Derq Teaches Driverless Cars to Safely Navigate Busy Intersections
There are many technical challenges to overcome in the development of self-driving cars, but one of the toughest is predicting and preventing collisions with pedestrians and other objects that end up in the roadway.
As urban cyclists know, there is a fair amount of intense eye contact that goes on between bikers and drivers as bikers seek to ensure drivers are paying attention—especially in the current era of constant smartphone distractions. That biker-driver interaction is something that can’t be easily replicated or taught to machines.
Derq, a Techstars Mobility alum based in Detroit and Dubai, is conducting a yearlong pilot project on Jefferson Avenue and Randolph Street in the Motor City to focus on vehicle- and pedestrian-related risks around intersections.
“We’re concentrating on road safety for regular cars as well as autonomous vehicles,” says Georges Aoude, Derq’s co-founder and CEO. “Our two main applications are intersection safety—predicting red light violations and sending warnings to a vehicle—and pedestrian safety, which involves leveraging cameras from smart cities to predict pedestrian intent.”
Derq’s software monitors vehicle-to-infrastructure and vehicle-to-pedestrian (V2X) interactions to predict and prevent collisions. The company is using patented artificial intelligence-driven algorithms, which Aoude developed while pursuing his aerospace engineering PhD at MIT, to create up to two additional seconds of warning time for vehicles and drivers.
“If you think of the Uber accident [which resulted in a fatality], they had the best sensors but the car still couldn’t see, or they didn’t work,” he notes. “We observe for a long time—on day one, we can’t make predictions. So we collect data and teach the vehicles what pedestrians are doing. The model is smart enough to adapt and predict intent, and we’re able to show we can do it with a high rate of accuracy and a low rate of false warnings.”
Aoude says Derq intentionally selected the Jefferson-Randolph intersection because it’s busy and complicated. It’s a major downtown thoroughfare connecting to highways as well as the route to access the Detroit-Windsor tunnel border crossing.
As part of the Detroit pilot project, Derq worked with the Michigan Department of Transportation (MDOT) to select and install sensors on the intersection. These sensors provide data feeds that allow Derq to run its software applications, generate predictions, and communicate threats in real time. (Derq chose FLIR Systems’ thermal imaging and combined visual/radar sensors for this deployment, Aoude says.)
“MDOT, the Michigan Economic Development Corporation, Techstars, and PlanetM made it possible for a startup like us to build our technology in Dubai and test it in Detroit,” he adds.
Derq was founded in 2016 and has raised a $1.5 million investment round so far. The company is in the process of seeking another round of capital in order to scale and enter new markets, Aoude says.
The Detroit pilot, which got underway in May, will run for a year, and the 10-person company hopes to eventually expand it to other MDOT intersections elsewhere in the state. Derq has also discussed further testing with the American Center for Mobility, located outside of Ann Arbor, MI.
“Our goal is to grow our presence to many other Michigan locations and grow our partnerships with players in the ecosystem,” Aoude says, adding that Derq already has signed 15 non-disclosure agreements with partners in the mobility industry as well as a memorandum of understanding and joint development agreement with Dubai’s government. “What we’re doing today can help [cars function better] now—we don’t have to wait for autonomous vehicles to be on the road.”