Road Rules: AI-controlled traffic signals

Question: Artificial Intelligence (AI) controlling stop lights seems much easier than AI driving a car. I’m not sure whether any are on the market yet, but seemingly AI stoplights could be an alternative to widening roads or installing new interchanges. And they would save fuel and help air quality by reducing vehicle stops and idling. When might we see AI stoplights in Washington?

Answer: So instead of a robot driving a car, you want robots to tell you how to drive your car? Actually, you’re totally right.

Developing AI traffic control signals is more doable than building fully self-driving cars. How do I know that? Because you can drive through cities with AI traffic control right now. One thing you can’t do right now is go buy a fully autonomous car. Yes, there are a few places you can ride in an AI taxi or shuttle, but those are essentially test vehicles limited to specific routes in a handful of cities across the U.S.

We’ve come a long way since the first traffic signal in 1868, a manually operated gas-lit fixture that was abandoned after it exploded and injured its operator. We’re all familiar with traffic signals now, but for most people their exposure to the behind-the-scenes management of traffic signals probably comes from all the movies where the signals get hacked.

Movies like The Italian Job (both versions), Hackers, Die Hard 4, Taxi (the French version) and others have turned signal hacking into a trope, but it also gives us a glimpse, however flawed, of how much technology is involved in getting through an intersection.

Over the last few years, AI experts have been testing out new systems for intersection management. Where traditional technology depends on engineers analyzing the data captured from cameras and sensors to make changes to traffic signals, AI signal management systems learn from the data on their own and make adjustments every second. Cities that have deployed AI intersection technology are seeing 25 to 40 percent reductions in travel time.

Maricopa County in Arizona has upgraded a lot of their intersections, and they estimate that if they swapped the gear in all 4,000 of their intersections for AI tech, it would save almost eight centuries of commuting every year. That also translates into less road-building and less pollution.

In terms of CO2, upgrading every intersection in the U.S. would be like removing 20 million combustion vehicles each year.

Smart intersections will become even more efficient as we get more smart cars on the road. Even if your car can’t completely drive for you, as more cars start talking with traffic lights your car will know when the light will change before you do. Intersections will also recognize and make adjustments for non-drivers, ensuring that pedestrians, especially folks with mobility challenges, have sufficient time to cross the street.

When will we see AI intersections in Washington? I don’t know of any cities using AI to improve traffic flow (If anyone reading this knows of some, let me know), but Bellevue has been using AI to increase safety by learning where the greatest risks are before a tragic crash happens. The technology is growing; I expect it’ll be here soon.

Despite names like Auto Pilot (Tesla) and ProPILOT (Nissan), all cars available to consumers require driver participation. We don’t have autonomous cars; we have cars with Advanced Driver Assistance Systems (ADAS). We’ve been dreaming of self-driving cars for decades, and they’re continually five to 10 years away, no matter what year it is. Meanwhile, AI is improving our driving in other less sensational but highly effective ways.

Doug Dahl is a manager with the Washington Traffic Safety Commission, Region 11 and publishes TheWiseDrive.com.