Site icon The Philadelphia Citizen

Ideas We Should Steal: Fix the Damn Potholes … with AI

A blurred partial view of the back of a man wearing a yellow safety jacket, holding a phone and taking a photo of a manhole for vialytics, an ai company that tracks potholes and other roadway problems.

Photo courtesy of vialytics.

In Philly, and most parts of the U.S., people report potholes predominantly through one method: call or email complaints to their city.

That system is … imperfect, at least in Philadelphia. When’s the last time you called 311 to report a pothole? And even if you did, a 2023 Axios analysis of Philly’s 311 data found that it takes the City 52 days on average to respond to pothole complaints — not the City’s promised three to five business days. This year, Philadelphia ranked 15 on the list of 50 worst cities for potholes.

Industry experts have estimated pothole repair costs the U.S. $2.5 to $3 billion annually — and that’s not counting payouts to drivers who’ve sued municipalities over pothole-induced accidents or vehicle damage. In Philadelphia, those payouts come to about $100,000 a year, according to a 2022 Philadelphia magazine analysis.

Some cities and townships across the U.S. and Europe are taking a more proactive — and promising — approach. Since 2018, more than 500 municipalities have partnered with vialytics, a German company that uses AI to assess road conditions. Here’s how it works: City workers — like members of Philly’s Streets Department, say — download the app onto their work phones, and put their phones on the dashboards of their city vehicles. Then, as they drive, an AI system assesses the road for potholes, mangled street signs and other kinds of damage. By 2030, vialytics estimates they’ll have mapped more than 248,548 miles of road — the distance from Earth to the moon.

“The status quo across the U.S. is what we call reactionary road management,” says Daniel Laufer, a vialytics senior sales manager whose territory is the Northeastern U.S. “We empower [cities] to start to identify damage to infrastructure and its early onset.”

This year, they’ve partnered with a number of cities in PA and New Jersey, including, recently, Conshohocken. Philly has experimented with using tech to detect potholes before, but a City program piloting similar tech ended in 2021 — without an ongoing contract. Perhaps it’s time we took another look?

Using AI to detect potholes and other pavement damage

Patrick Glaser, Achim Hoth and Danilo Jovicic-Albrecht founded vialytics in Stuttgart, Germany a little more than five years ago. An engineer, app developer and entrepreneur, respectively, the trio wanted to solve the dual problems of poor roadways and lack of city workers to help monitor road conditions and repair pavement. AI seemed like it would do a decent job identifying damaged roadways; tech could reduce labor needed to do that work.

“The more that we can start to become proactive in road management instead of reactive, the more we start to be able to … produce safer and smoother roads for citizens.” — Vialytics’ Daniel Laufer

So, they designed an app that takes videos of roadways and assesses them for 15 different kinds of pavement distress, including potholes, sinkholes, alligator cracking and damaged road signs. Once installed onto a phone, vialytics detects and logs locations of problems to create a map engineers can use to identify and address damages. The system also has a feature that allows engineers to assign different teams to repair the damage detected by the app. Vialytics expanded to the U.S. in 2023.

Laufer says the system is both more accurate and faster than the traditional methods cities use. “Typically you would hire an external engineering company to come in and do visual inspection, and you’re getting a subjective assessment,” he says. “Let’s say you have 100 miles of road. It could take anywhere from three to six weeks. Vialytics can cover that in a day … and we’re also getting an objective rating.”

Edison, New Jersey, population 106,836, with 307 miles of roadways, is a client; their public works department reports that the tool has saved them 20 hours per week. (Philadelphia, for comparison, has a population of 1.6 million and has 2,525 miles of roadways, although some, like our highways, are maintained by the state through PennDOT.)

Why use AI to detect potholes?

A lot of cities rely on resident complaints to identify potholes, perhaps foolishly. Most people are pretty bad at identifying the difference between a pothole, which the City of Philadelphia defines as a divet less than 10 inches deep, and the much larger sinkhole, which are craters more than 10 inches deep, or other kinds of pavement damage. Vialytics separates potholes into six distinct categories, so street department workers know what they need to address most urgently.

“The more that we can start to become proactive in road management instead of reactive, the more we start to be able to … produce safer and smoother roads for citizens,” Laufer says.

Collecting data on road conditions can also help cities ensure they’re acting equitably when they decide what streets need improvements. People in more affluent parts of a city tend to have more time to call in and advocate for their streets; when they complain, they receive faster responses.

Photo courtesy of vialytics.

“Oftentimes it’s the squeaky wheel that gets the grease — which residents are most upset about their road,” says James Gant, borough manager for the Borough of Red Bank, New Jersey. Red Bank started using vialytics in April, when they signed a 36-month, $14,000 contract with the service. Laufer says cities pay an annual subscription fee for the service that varies based on the size of a city and the amount of roadway it has.

“This gives us a good baseline and something that we can show the community to explain why we are picking certain roads over others,” says Gant. “A lot of times people do feel like it’s arbitrary, and it shouldn’t be. It should be something that is actually done with some sort of analytical backing to it.”

Saving time, improving response

For many vialytics clients, the primary benefit of the tech comes from allowing a municipality to proactively respond to potholes, rather than waiting for residents to complain — or worse, request the city pay for a vehicle damaged by poorly paved roadways. Potholes cost on average about $25 to fix. But they cost a lot more if you don’t fix them. In a typical year, the City receives between about 150 to 175 auto damage claims from residents related to potholes. In 2022, the City paid an average of $575 per claim, per Philly mag.

Chambersburg, PA, began using vialytics in May. The borough owns all of its utilities, so it often has to dig under and patch roadways to improve gas, water and other utility lines. They wanted a service that could monitor repairs and patches, as well as detect new pavement damage. Their sanitation, engineering and highway vehicles are equipped with the app and drivers let it run while they conduct other public works business. Their three-year contract set them back around $13,000.

“We’ve always kept very good track of our roadways. We’re just trying to do it a little more efficiently,” says Brian Fogal, the borough’s engineering supervisor.

Chambersburg and Red Bank both said it’s too soon to see how vialtyics has reduced their annual pothole count. Additionally, discerning reductions can be tricky. Cities that rely on civilian reporting systems don’t always have good data on the number of actual problems. A pothole could sit for weeks before someone bothered to call it in. Plus, tough winter weather tends to increase potholes.

“This gives us something that we can show to explain why we are picking certain roads over others. A lot of times people feel like it’s arbitrary, and it shouldn’t be. It should be something that is actually done with some sort of analytical backing to it.”  — Red Banks’ James Gant

What vialytics has done is improve the response time. It used to take Chambersburg four days to respond to a pothole complaint. Now, most are filled within a day or two of being identified. And, the app has reduced the time it took for the city’s public works engineers to map problems along all of the 7-square-mile borough’s roadways from months to a week. Grant says Red Bank has identified 230 different pavement issues since starting with vialytics in April.

“We’re able to collect data year round,” says Charles “Chuck” Nipe, director of public works for the Borough of Chambersburg. “It provides a very objective analysis of your road network and street signage.”

How could Philly use tech to improve pothole response?

Philly, with its nearly 2,000 miles of paved roadway, is obviously much bigger than either Red Bank or Chambersburg. Laufer recently moved to Queen Village from New York City called Philly “its own pothole ecosystem.” He compared our roads to coral reefs, because they’re unique and because of their age, materials (Belgian block, cobblestones, or, most typically, paved-over brick) and 300-plus years of traffic. Seasonal temperature fluctuations — especially snow and ice — exacerbate issues.

But other big cities are using these kinds of tools. Memphis, TN has worked with Google and the machine learning company SpringML, on a similar program. They project the AI-detection software will lead to 75 percent more potholes being identified. In Europe, where vialytics has been around longer, cities like Laichingen, Germany, report that they can completely map their streets in six days now, compared to a previous six weeks. The 60,000-person city of Schwäbisch Gmünd, another German client, reports they’ve fixed 800 potholes in the three years they’ve used vialytics. Their road grade went from 3.35 to 2.94. (Higher grades indicate rougher roads.)

Nipe and Fogal appreciate how proactive vialytics customer service is — that’s one of the things that stands out to them about the product. They have calls with vialytics support every two weeks where they can get help with the service and provide feedback on the product.

“If we have a question, they respond to it. If we have input, they consider it with their development team,” Nipe says.

Whether Philadelphia partners with vialytics or a different AI pothole detection provider will mean little, however, if the City doesn’t do anything with the data it collects. In 2020, the City partnered with a similar company, the North Carolina-based GoodRoads, on a pilot program that used sensors to assess the condition of pavement throughout Philly. The pilot, which the Streets Department and the Smart Cities team conducted in tandem, assessed 1,100 miles of Philly streets — about half of the city — in 438 hours of driving time between 2020 and 2021. That data was used to update the pavement condition index, a tool used by the Streets Department to determine which roads to prioritize for resurfacing projects.

But Streets did not opt to continue working with GoodRoads when the pilot ended. Nor would the department share data from the program. A spokesperson says they plan to continue exploring new technologies to help improve road conditions, without announcing any specific plans or timelines.

In other words, Philly seems to have “concepts of a plan” — which, sadly, is pretty much the norm once the City’s tax-funded pilot programs conclude. Even if pilots show promise — like the focused deterrence policing Group Violence Intervention pilot — they either stay small, never scale, or disappear.

Laufer says vialytics would be open to working with the City of Philadelphia, and that the company has been in discussions on a state level with PennDOT. He notes that working with a large city requires significant resources — from the City and vialytics alike — and sometimes starting with a pilot in a portion of a city, say, several neighborhoods where road conditions are more representative of the problems the city faces, can be beneficial.

“Implementing change at that level, in a city of this size, requires significant resources,” he says. “But it’s a very easy-to-use technology that can be scaled, whether you have three users or whether you have 100 users.”

MORE IDEAS WE SHOULD STEAL

Exit mobile version