Introducing Crowdsourced Pavement Quality Maps
Lvl5 Releases List of States with Best and Worst Road Quality Based on 5 Million Miles of Payver Driver Data
tl;dr: Michigan has the worst roads in the U.S. while Florida has the best. You can contribute to the ranking database by using the free dashcam app, Payver.
October 30, 2018 — San Francisco, CA — Lvl5, a company that creates HD maps for self-driving cars, today released a list of states with the best and worst road quality. The company captured and analyzed five million miles of driving on U.S. roads and found that Michigan has the worst roads in the country while Florida has the best.
State and federal governments spend over $400 billion on maintaining and building new roads each year. As US infrastructure deteriorates by the day, taxpayers are left wondering if initiatives such as the White House’s proposed $1.5 trillion infrastructure spending will be used efficiently to bring relief.
One of the major obstacles cities face in maintaining their roads is that their data is out of date. Cities contract out to firms to map their roads using LiDAR, which is both expensive and slow. Many cities, including Los Angeles, only receive this data once per year, meaning the city depends on months-old data to fix roads today. Lvl5 aims to help cities more efficiently maintain their roads by giving them up-to-date data that enables them to quickly identify road problems (e.g. crack in a road) before they become larger and more expensive issues (e.g. a pothole).
Lvl5 captures hundreds of thousands of miles of video data every month through their iPhone dashcam app, Payver. The app pays users up to $0.05 per mile to record their driving using their cell phone. The company is already working with many Uber and Lyft drivers across the US who run the dashcam app throughout the day. As cars are driving around, they are “vacuuming” up data about the roads. The company then uses its computer vision algorithm to translate that crowdsourced footage into maps which rate road quality and show cities where problems exist.
The country-wide dataset spans over five million miles of Payver driving data from the last year. The methodology for ranking these videos included randomly selecting video frames from the videos (in total, 15 million frames sampled), and excluding all but the surface of the road from predictions by a neural network. Their neural network measures quality in four distinct areas: road paint fading, pavement cracking, potholes, and surface flatness. The data was normalized by frame density, and filtered to remove areas with not enough data.
The company also analyzed if higher gasoline taxes or higher spending on construction impacted road quality, and found no correlation for either.
How to Obtain Road Quality Data For Your City
Please contact lvl5 sales: email@example.com, lvl5 is happy to demo the technology in your area.