IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Don't Fix the Worst First: How Next-Gen Tech Tools are Changing Road Maintenance

Two solutions using data analytics and artificial intelligence may serve as a model for how cities across the country can take better care of their streets.

As city populations nationwide continue to grow, more residents move out of urban cores and into peripheral suburbs, causing an increase in commutes (PDF) — which puts more strain on already-decaying infrastructure. While road maintenance is something every city, town and state must consistently deal with, some cities are employing tech-based solutions to prevent the problem from happening in the first place.

Generally cities act to patch and repair potholes depending on their severity and public nuisance. Fixing a large hole in the middle of a thoroughfare, for instance, will likely be prioritized. Smaller potholes on less dense residential streets, however, are generally put on the back burner, creating a backlog. This scenario is often referred to as fixing the worst first; the emergency repairs often drive up costs in the long run as they are temporary replacements that will ultimately need replacing down the line.

RoadBotics, a startup out of Carnegie Mellon University, focuses on monitoring the status of roads before emergency crews are needed to patch and repair. Users mount smartphones against the vehicle’s windshield with the camera facing out, drive the roads up and down, and gather detailed video of street conditions and signage. The camera, which captures 30 frames per second, provides more than enough information necessary for road maintenance crews and public works departments to identify and monitor which streets will need maintenance, explained CEO Mark DeSantis.

Last week, DeSantis surveyed the roads of Pittsburg, Calif., which lies roughly 40 miles east of San Francisco. Guided up and down each street by Associate Product Manager Helen Meng, the RoadBotics team worked to drive every road, creating a continuous video log.

The video and accelerometer data is fed into an AI-powered algorithm that automatically classifies 3-meter segments of road into a system that's color coded depending on the damage to the street. Broken down into green, yellow and red segments depending on cracks, potholes or debris, users can then access a map view with streets color-coded to see where the problematic areas are.

“It’s all about giving the user actionable information,” said DeSantis, adding that while it provides an image and GPS location of sections that need immediate attention, “the system is really about targeting those yellow [moderately damaged] roads. If we are able to catch the damage before it needs immediate repair, we can salvage the road.” This method has the potential for saving municipalities hundreds of dollars, he said.

While RoadBotics is currently gathering information in partner cities themselves, there is enormous potential in outfitting existing city vehicles with these smartphones to continuously gather information on roadway conditions. The data is owned by whoever collects it, however, DeSantis said that even if RoadBotics owns the information, it would never be distributed or sold without consent from the municipality.

DeSantis also lauded how international cities handle their road maintenance efforts. “They focus more on monitoring rather than repairs. The key [for U.S. cities] is not to spend more money, it's simply to monitor roads more frequently.” RoadBotics has partnerships with North Huntington, Pa., and in Pittsburg, working with the Contra Costa Transportation Authority, as well as international programs in Berlin, Tokyo and Melbourne, Australia.

“Traditionally there have been two methods for monitoring roadway conditions,” DeSantis said. The first, “folks in a truck drive around usually once a year and identify features by hand.” This process is not only tedious, but also incredibly subjective. The alternative is renting or purchasing “super-expensive vans outfitted with sensors, cameras and lidar,” which is cost-prohibitive for many cities. DeSantis hopes that RoadBotics provides a third, more manageable option.

Bob Bennett, chief innovation officer of Kansas City, Mo., is providing his city with another option. In trying to deal with the omnipresent pothole problem, Bennett and the rest of the city is looking inward at data already being collected by the city to predict where potholes will occur.

Kansas City is already collecting mountains of data through its Smart City Program. The pothole predictor, which combines the efforts of the city and Xaqt, to look at multiple variables including “precipitation, temperature change day to day, volume of traffic, types of vehicular traffic, and condition of the roads as they currently exist,” explained Bennett.

“America’s infrastructure is old. It's decaying,” he said. “That bill is not going away, if anything it's going to increase. What this technology allows me to do is minimize that increase.”

The city ran the algorithm against road data from 2003 to 2012, and was able to predict where potholes formed with 76 percent accuracy. Bennett decided to move forward and begin a pilot on six major streets, using more data that is being collected now, which he expects will raise the accuracy close to 95 percent and identifying areas within a 5- to 10-foot radius.

Looking at preventive maintenance, the city would be able to catch and repair problematic features before they become a headache. “By doing this as routine maintenance, we’re able to use materials and equipment we already have allocated for those tasks,” said Bennett. “We’re just doing it more intelligently now.”

And as DeSantis points out, "We are at a point where potholes are completely unnecessary. The reality is, if cities spent the time monitoring the road, they wouldn't need to rehab roads.”

Ryan McCauley was a staff writer for Government Technology magazine from October 2016 through July 2017, and previously served as the publication's editorial assistant.