How Artificial Intelligence Is Cutting Wait Time at Red Lights

Who hasn’t been stuck seething at an interminable red light with zero cross traffic? When this happened one time too many to Uriel Katz, he co-founded Israel-based, Palo Alto, California–headquartered tech startup NoTraffic in 2017. The company claims its cloud- and artificial-intelligence-based traffic control system can halve rush-hour times in dense urban areas, reduce annual CO2 emissions by a half-billion tons in places like Phoenix/Maricopa County, and slash transportation budgets by 70 percent. That sounded mighty free-lunchy, so I got NoTraffic’s VP of strategic partnerships, Tom Cooper, on the phone.

Here’s how it works: Sensors perceive, identify, and analyze all traffic approaching each intersection, sharing data to the cloud. Here light timing and traffic flow is adjusted continuously, prioritizing commuting patterns, emergency and evacuation traffic, a temporary parade of bicycles—whatever. Judicious allocation of “green time” means no green or walk-signal time gets wasted.

I assumed such features had long since evolved from the tape-drive traffic control system Michael Cain’s team sabotaged in Rome to pull off The Italian Job in 1969. Turns out that while most such systems’ electronics have evolved, their central intelligence and situational adaptability have not.

Intersections that employ traffic-sensing pavement loops, video cameras, or devices that enable emergency vehicle prioritization still typically rely on hourly traffic-flow predictions for timing. When legacy system suppliers like Siemens offer similar technology with centralized control, it typically requires costly installation of fiber-optic or other wired-network connections, as the latency inherent in cellular communications can’t meet stringent standards set by Advance Transportation Controller (ATC), National Electrical Manufacturers Association (NEMA), CalTrans, and others for safety and conflict resolution.

By contrast, NoTraffic localizes all the safety-critical decision-making at the intersection, with a camera/radar sensor that can identify vehicles, pedestrians, and bikers observing each approach. These sensors are wired to a box inside the existing control cabinet that can also accept input signals from pressure loops or other existing infrastructure. The controller only requires AC power. It connects to the cloud via 4G/5G/LTE, but this connection merely allows for sharing of data that constantly tailors the signal timing of nearby intersections. This is not nano-second, fiber-optic-speed critical info. NoTraffic promises to instantly leapfrog legacy intersections to state-of-the-art intelligence, safety sensing, and connectivity.

Installation cost per intersection roughly equals the cost budgeted for maintaining and repairing today’s inductive loops and camera intersections every five years, but the NoTraffic gear allegedly lasts longer and is upgradable over the air. This accounts for that 70 percent cost savings.

NoTraffic’s congestion-reduction claims don’t require vehicle-to-infrastructure communications or Waze/Google/Apple Maps integration, but adding such features via over-the-air upgrades promises to further improve future traffic flow.

Hardening the system against Italian Job–like traffic system hacks is essential, so each control box is electrically isolated and firewalled. All input signals from the local sensors are fully encrypted. Ditto all cloud communications.

NoTraffic gear is up and running in California, Arizona, and on the East Coast, and the company plans to be in 41 markets by the end of 2021. Maricopa County has the greatest number of NoTraffic intersections, and projections indicate equipping all 4,000 signals in the area would save 7.8 centuries of wasted commuting time per year, valued at $1.2 billion in economic impact. Reducing that much idling time would save 531,929 tons of CO2 emissions—akin to taking 115,647 combustion-engine vehicles off the road. The company targets jurisdictions covering 80 percent of the nation’s 320,000 traffic signals, noting that converting the entire U.S. traffic system could reduce CO2 by as much as removing 20 million combustion vehicles each year.

I fret that despite its obvious advantages, greedy municipalities might push to leverage NoTraffic cameras for red light enforcement, but Cooper noted the company’s clients are traffic operations departments, which are not tasked with revenue generation. NoTraffic is neither conceived nor enabled to be an enforcement tool. Let’s hope the system proves equally hackproof to government “revenuers” and gold thieves alike.

The post How Artificial Intelligence Is Cutting Wait Time at Red Lights appeared first on MotorTrend.