transportation

U.S. Drivers Have Lost $8 Billion to Red Lights — Can AI Traffic Signals Save Us?

This company is using AI to manage traffic signals to reduce the time drivers spend idling at red lights, cutting emissions, and saving money.

The Department of Energy says idling for more than ten seconds burns more fuel and produces more CO-2 than shutting off and restarting your engine.

A 2021 traffic signal report says drivers in Nevada spend more time idling at red lights than drivers in any other state. Nationwide, Americans spend 6.8 percent of each trip at red lights, accounting for more than $8 billion in lost time, energy, and fuel. The map below shows how red light delays compare across all 50 states.

More From LX News

The Pandemic Made Us Drive More Recklessly in a Deadly 2020 on the Road

Tim Menard says artificial intelligence can unclog our roads and reduce red light idling. He started LYT, a cloud-based platform that integrates drivers’ navigation data (which he says is anonymous and collected passively) with a city’s network of traffic signals, analyzes traffic flow, and adjusts traffic light timing accordingly. His tech is currently in use in cities in California, Oregon, and Washington. I recently caught up with Menard to talk about traffic management and smart traffic lights.

This interview has been lightly edited for brevity and clarity.

Peter Hull: Traffic jams were happening long before automobiles came along. And roads and vehicles have changed over centuries. Given all this history, what’s your take on the state of traffic management today?

Tim Menard: You bring up a good point that traffic's always been around. We've had roads for hundreds of years and they've seen evolution over long periods of time. Only up to the last hundred years has that really accelerated decade over decade, to our last decade which we've now seen with bikes, scooters, cars, and machinery. We're talking about autonomous vehicles, we're talking about drones. So we have the same structure we've had for a hundred years, but now we've got six, seven different types of mobility options, all competing for that same couple-lane wide space. And it just cascades quickly.

PH: At LYT you design tools that make traffic signals smarter. How does this improve traffic flow for everyone?

TM: Those who want to participate in their drive, they’re going to Waze, trying to look at “how can I try and game the system? What can I take, what other routes?” We've seen how that's provided 2 to 5 percent of travel time improvements for yourself. But what if all that information from navigation was able to get back to the actual traffic management system? That's why the perspective in AI is so interesting, because, with a bird's eye view and this new understanding of, not trying to detect and react, but switching to a knowing and a treating, you now switch the gains to everybody being optimized. 

And there are some phenomenal academic papers that have been able to show that you can take our infrastructure and provide everybody [with] a 30 to 40 percent improvement in total regional transportation. I imagine going to any major city right now — Chicago, Los Angeles, New York — and being able to cross from one boundary line to the other 40 percent faster. It's a night and day difference and game-changer for the entire economy of that city. And that's what we're after.

PH: So by sharing a bit of travel data, you’re able to help optimize the road for everyone. What risks are there for people who share travel data?

TM: We've seen a lot of different technology applications that have really gone deep into trying to understand who you are. But when it comes to a mobility situation, the philosophy we've taken is that if you're looking at these major cities, that's a lot of data. If you have eight million people that’s eight million data points essentially, and they're accounting for that every second. What makes us more effective and efficient is [that] it doesn't actually matter who you are. What matters is just having the small amount of information that (a) you're on the road and (b) you're a participant. It's the aggregate whole; being able to share small amounts of data to tell the big picture. And when you do that, you can remove the privacy concern because it's truly autonomous and it's just "where are the cars," "where are the people," "where are the trucks," "how did they move on this day?"

PH: We’d probably all like to spend less time idling at red lights. Does your platform address emissions?

TM: So we've all heard over the years how bad stop-and-go traffic is, especially on the internal combustion engine where you're putting more fuel in, then all of a sudden you hit the brakes so now you've got brake dust

On just a “my vehicle your vehicle” basis, this system has got the potential that if you're making improvements to the whole region, you’re dropping those emissions, you're dropping fuel usage by everybody, and your vehicles are running effectively and efficiently and that's got a huge takeaway. Where we've been able to demonstrate this is reducing emissions with [bus] transit. You can imagine: they're running mostly 16 hours a day. They're going on all arterial, side streets, going through neighborhoods, constantly stopping and going. So just for them [buses], maybe they hit that same red light every day, but that bus route might have 50 buses, 365 days a year. Those brakes add up and emissions around that certainly add up. In San Jose, we're able to show that by getting intersection delays next to zero, that wound up being a reduction of 12 percent on emissions and brought back fuel economy. So savings actually paid for the whole system itself in about 18 months.

PH: People were afraid of the first traffic signal in London and today people are afraid of AI. How do you address concerns about giving a machine control over our traffic systems?

TM: So A.I. is a full spectrum. We've seen the movies. The good news is we're not making a robot. We're confined A.I.; it's very specific, and it's actually just a part of an entire chain of processes. So if the AI is making weird predictions or not giving results that are within reason then it's just kind of caught and junked right there and it doesn't go anywhere else. So there's a lot of interlocks. AI gets thrown around a lot, [but] it's where is the AI and how much control does it have? In a setup here where it's very systematically controlled, there's a lot of fact-checking that you can do on information and results.