The Pittsburgh company is adding new features and more granular results to its software for analyzing pavement damage, with plans to move into larger cities and smaller counties in the future.
The latest development from RoadBotics, announced earlier this month, is the ability of its pavement-assessment software to flag and catalogue roadway damage, or “distresses,” into specific categories: potholes, patches and sealed cracks, fatigue cracks, longitudinal and transverse cracks, pavement distortions and surface deterioration.
Up to this point, municipal and county governments have been able to use the RoadBotics app by putting a smartphone in a windshield, camera forward, and letting it collect image data of the jurisdiction’s roadways. The app would then upload those images to the RoadBotics platform, where deep-learning technology would analyze them and assign a rating for every 10-foot section of road on a scale of 1-5, measuring the quality of the pavement. The app would then send the results to an interactive online mapping platform called RoadWay.
Vice President of Growth Ryan Gayman explained that with the addition of the new distress identification feature, a user can click a checkbox next to “potholes,” for example, and the map will light up like a Christmas tree showing all the potholes, in addition to the 1-5 rating for each 10-foot segment of road. Gayman described this as a major hurdle cleared, the first in an impending series of new features that will only be possible because of the vast amounts of data the company has been collecting and using to improve its machine learning engine since its launch in 2016.
“This has been the buildup of our company — being able to continue to collect, process, find those edge cases, and develop a library of labeled data that enables us to do this,” he said. “It’s taken years and millions of dollars to get to this point, so it’s really a massive leap … moving from that 1-5 condition score to what folks have been asking us for, for quite some time, which is: show me where my potholes are, show me where my longitudinal cracks are, so on and so forth.”
Gayman said the company envisioned distress identification as a tool for government agencies who need to start specific maintenance, like a crack-sealing program or filling potholes, but can’t afford to do it mile by mile. He was adamant that the app is a decision support tool, not a decision-making one; it doesn’t tell customers what they should do but it makes maintenance easier to plan and prioritize.
“More granular data ties directly to them being able to make maintenance decisions,” he said. “Our hope is that it enables people to do more preventive maintenance measures … We want to provide increasingly valuable data and insights to them so that they can interpret that for their own planning purposes.”
Gayman said the company is working with more than 200 clients in 15 countries and branching into new customer bases. Since RoadBotics spun out of Carnegie Melon University four years ago, it has raised $11.4 million from urban tech investors such as Urban.Us, URBAN-X, Radical Ventures and Ekistic Ventures, according to a news release, and served predominantly medium-sized cities like Savannah, Ga., and South Bend, Ind. Gayman pointed out that those have more flexibility than some larger cities in adopting new technology and methods, but he said RoadBotics is starting to see interest from larger metros, too. Its largest client to date is the city of Detroit, with 2,600 centerline miles, and the company intends to pursue contracts with more of those and even state departments of transportation in the near future. He said small county governments are also of interest to the company, because they often have small budgets to maintain long roads spread out over many acres.
Four years into the game, RoadBotics doesn’t appear to have many direct competitors. Gayman said there are some mobile apps that use accelerometer data and vibrations to create a roughness index for assessing road quality, but those only pick up road artifacts that go under the vehicle’s wheels. Other than that, he said the company has developed an enterprise partner program for licensing its technology to private civil engineering firms, which use RoadBotics to deliver better services to the governments they work with, but in that sense they’re not true competitors.
He said there are also companies that use lidar or ground-penetrating radar to assess roads, but some of those use RoadBotics as a first pass and then deploy more specific tools based on what they find.
Gayman said up next for RoadBotics will be more detailed tools, ways to annotate and input data and record work done on roads after the fact. He said the company also wants to assess other features in the road besides damage, such as lane markings, which are already being read by autonomous vehicles.
“The way those vehicles are navigating, and ensuring there’s not an accident by lane-wavering, is by having high-quality things like line-striping,” he said. “Because we’re capturing a full, high-definition image, we can look at the other things that a public works director … and ultimately the people who have to experience that asset, care about on any given day.”