Today, we take the safety and order created by sidewalks for granted, but our roads are still far from perfect — traffic accidents claim 1.25 million lives each year. That’s why Eran Shir is working to build his own version of the sidewalk for drivers. It’s called Nexar, an AI-based dashcam that allows cars to communicate with one another fluidly enough to avoid collisions. His goal is to create real-time virtual maps that guide cars around potholes, over icy roads and past stalled trucks like a perfectly played video game. As Shir, who co-founded the Tel Aviv-based company in 2014, puts it: “We envision becoming a building block in the air traffic control system for the road.”
Shir isn’t the only one with this vision. For the last two decades, both Detroit and Silicon Valley have been on the prowl for advanced driver-assistance systems (ADAS) — smart technology for developing autonomous vehicles (AV) and an infrastructure to prevent them from crashing into one another. Last March, when Intel acquired AV-technology company Mobileye for a record-breaking $15.3 billion, the chipmaker estimated the ADAS industry would grow to $70 billion by 2030.
Yet amid the self-driving-car hype, Nexar has spotted an undiscovered twist — one ingenious enough to attract $30 million in funding from investors, including GE Ventures. Rather than embedding technology in a clever chip or sleek device, Nexar relies solely on the hardware and sensors sitting in your pocket right now: the ubiquitous smartphone.
Most drivers own a smartphone, and with each one added to its network, Nexar grows a little wiser about the roads and the cars on them. So far, its users hail from 740 cities in 160 countries and drove over 100 million miles last year. They added 10 million of those miles in December alone. When it comes to data, Nexar is stinking rich.
Here’s how it works: Once downloaded and linked to a driver’s smartphone camera or dashcam, the app puts all smartphone sensors on the lookout for danger. Nexar runs its AI algorithms on the camera feed to detect vehicles, traffic lights and other objects around the vehicle. It uses the sensors to determine the vehicle location and trajectory. For example, the compass can be used to detect power lines by monitoring fluctuations, while the gyroscope picks up on potholes below. Nexar’s algorithms crunch this phone data to determine the safest driver response and the app swiftly transmits the message to other Nexar-powered vehicles nearby via the cloud. This crowdsourced vehicle-to-vehicle network gives drivers what Shir calls “superpowers,” e.g., the ability to look around the corner or see 10 blocks ahead. If, say, you’re cruising up Broadway in Manhattan, Nexar’s network may warn that a steam pipe erupted several blocks to the north — prepare to stop suddenly. All of this takes place within 5 to 10 seconds.
It’s no wonder 10 percent of ride-share drivers subscribe to Nexar in New York City, where its network can scan each Manhattan intersection every 5 minutes. A new service there can alert fleet drivers whenever they’re near a garbage truck or school bus, vehicles whose stop-and-go movements notoriously gum up traffic. So far, Nexar says, its users in New York have reduced collisions by 30 percent.
Last year, as part of a partnership with a Nevada agency to generate 250 million miles' worth of road data each week until 2022, Nexar noticed that drivers in Las Vegas would often slam on their brakes or abruptly hit the gas on a series of backroads — bizarre behavior even by Vegas standards. As it turned out, those drivers were responding to swarms of gamblers using lesser-known alleyways to move between casinos, a safety hazard that the city can resolve with new regulations or perhaps better road design.
Shir believes Nexar could provide similar insights to companies that make autonomous vehicles, which rely on a battery of algorithms to protect passengers. Last year, its network collected data on 11 million driving incidents as varied as run-of-the-mill sideswipes and cows stuck in the road — dangerous scenarios that product developers can use to test AV performance and create algorithms that teach vehicles how to respond best. The more scenarios an AV’s computer anticipates — and can learn from — the safer everyone on the road will be.
In the meantime, Shir is still waiting to see how this air traffic control system for the road takes form. “I think it will come from an angle that is not obvious,” he ruminates. “It’s a nonobvious proposition.” Perhaps the solution is as simple as the sidewalk — or maybe a cellphone.