AI Cuts Crashes 17 Percent in Vegas — But There's a Catch

The experiment, which used AI to identify hot spots that police and transportation officials then responded to, was limited to one section of highway. And there are other mitigating factors — like hockey, for one.

by / November 26, 2018

A traffic crash-predicting startup, Waycare, has successfully demonstrated that its artificial intelligence — paired with specific responses from law enforcement and transportation officials — can reduce highway collisions.

But the company still has a long way to go in proving itself. It's not yet clear how its success metrics would change under different circumstances.

The experiment, as a topline, cut crashes 17 percent on a stretch a couple miles long on Interstate 15 where it meets Russell Road near the iconic “Welcome to Fabulous Las Vegas” sign.

But there are a lot of asterisks on that statement. The math works out like this: From May through July this year, that section of highway saw 57 crashes in a span of 92 days. From August through September, when Waycare was running its program along with its state and local government partners, there were 15 crashes in a span of 29 days. The number of primary crashes per day dropped from 0.62 to 0.52, a drop of about 17 percent.

That’s as opposed to secondary crashes, which is when a slowdown caused by one accident causes another crash.

Observation period Number of days in observation period Number of crashes Crashes per day
May-July 92 57 0.619565217
August-September 29 15 0.517241379

Waycare simply supplied the data and analytics; the interventions that actually contributed to the decrease came from a collaboration between the Nevada Highway Patrol, the state Department of Transportation and the Regional Transportation Commission (RTC) of Southern Nevada.

How did they do it? Basically, they got drivers to slow down in the right place at the right time.

Using different kinds of data, Waycare’s AI came up with predictions about when and where crashes were likely to happen. Then the three government partners stationed police vehicles in those areas with lights flashing and put up messages on digital highway signs telling drivers to reduce their speed.

That’s because, as Waycare CEO and co-founder Noam Maital puts it, of the “snaking effect.” As traffic slows down, a car brakes to avoid hitting the next car ahead. And then the car behind that one brakes, and so on.

What makes the intervention work is this simple truth: The faster a car is going when it applies the brakes, the longer the distance — and the more time — it will need to slow down. Reduce the speed, reduce the braking distance and time.

"Because of that shorter period, when you take that and multiply that by thousands of vehicles, the throughput on the road improves," Maital explained.

What the group found was that not only did the entire traffic jam become more efficient at clearing itself, but there were fewer primary accidents as well. And because of the data collection, the transportation officials were identifying accidents about 12 minutes faster than usual. And because those highway patrol officers were already stationed in high-risk areas, when there was a crash they were often able to get there faster.

“The quicker we can respond, the quicker we can clear lanes, the better everything is,” said Brian Hoeft, director of RTC’s Freeway and Arterial System of Transportation division.

The 17 percent reduction only counts primary crashes, because measuring the reduction in secondary accidents is trickier, Maital said. But since a secondary crash can only happen after a primary crash, it’s likely that the experiment avoided secondary crashes as well.

That’s not the only asterisk next to that 17 percent statistic. The test period (August and September) was considerably shorter than the control period (May through July), leaving space open for the possibility that a longer test period might have yielded different results. There’s a huge construction project — Project Neon — not too far away, and changes in construction can change traffic patterns.

There’s also the matter of the Stanley Cup, the annual championship series of the National Hockey League, which pitted the Las Vegas Golden Knights against the Washington Capitals. Three games were played at the Golden Knights’ stadium, which sits just off I-15 not far from the corridor Waycare tested on. Those games happened in late May and early June, during the control period Waycare was measuring against. It’s possible that increased traffic during those times could have increased crashes, making it look like the collision rate dropped farther than it actually did.

But Maital, Hoeft and Waycare CIO Shai Suzan all remain confident that the program did, in fact, reduce collisions, even if the experiment hasn’t gone through the scientifically rigorous process that articles in peer-reviewed journals do.

“We are confident that the program plays a big role in causing that reduction, and we do acknowledge the potential asterisks, the other stuff going on,” Hoeft said.

One big reason for that confidence is that any drop in collisions runs counter to the trends that envelop transportation in Las Vegas and Nevada. Statewide, Maital said accidents have been increasing.

And on the entire I-15 corridor — not just the segment involved in the experiment — there was virtually no change in the number of crashes per day during the same two time periods in 2017, according to Suzan.

The company will be able to give better insight next year, since it will continue monitoring that stretch of I-15. At that time it will be able to provide something closer to an apples-to-apples comparison to show how the experiment truly impacted crash rates.

“We’re wringing out these numbers and evaluating them and getting more and more reports ready and as time moves on we’ll be able to do it from a more seasonal basis and step it up,” Hoeft said.

In the meantime, the partners are considering different avenues for the future. Could Waycare be expanded to cover more stretches of highway? Surface streets? Could lane-by-lane signs help change driver behavior in even more specific ways to avoid crashes?

Or, looking even farther ahead, what about self-driving vehicles?

"In a connected and autonomous environment, these types of things can be automated to include speed reduction as well," Maital said.

Ben Miller Associate Editor of GT Data and Business

Ben Miller is the associate editor of data and business for Government Technology. His reporting experience includes breaking news, business, community features and technical subjects. He holds a Bachelor’s degree in journalism from the Reynolds School of Journalism at the University of Nevada, Reno, and lives in Sacramento, Calif.