LVL5’s “Pavdar” system – recording smartphone video from dashboards – may be the future of pothole detection. Already, it knows Michigan roads are the worst.
It’s very hard to get a national, state or even citywide count of potholes on our streets and highways. Which is problematic in two ways: To repair them, first you have to know where they are. Second, to budget for those repairs, you have to know how many and how bad they are.
Google Satellite hasn’t yet managed to provide a count because the technology isn’t able to get that much information on the granular (potholes and pavement cracks) level (Pothole.info gets asked about this a lot). But there are some technologies in use that attempt to count potholes: Lidar – which uses laser pulses from a puck-like, roof mounted mechanism (as previously reported on Pothole.info) – is being used by Uber, Toyota and other companies that are planning for the autonomous vehicle era.
With better detection of pavement problems those cars will operate with greater efficiency – and fewer flat tires. A single Lidar unit costs about $4,000 per set, which limits its use and adds to the cost of autonomous vehicles.
Pavdar can see the potholes – at a lower cost per car
But another technology is being deployed at a much lower price. It is called Pavdar, produced by a company called LVL5 (pronounced “level five”). It operates on smartphone cameras, mounted on car dashboards in such a way that the camera records an oblique-angle view of the roadway in front of the vehicle.
Pavdar is a new application of crowdsourcing, enlisting regular vehicle drivers in the quest to identify road imperfections like potholes and pavement barriers (it looks for other things as well, such as bollards, temporary obstructions, traffic-calming cutouts, etc.). Those drivers mount their own smartphones to face forward, turning their phones into dashcams that can see what’s there, record it, and then transmit it to the central database. There a computer vision algorithm translates the footage into maps that rate road quality, pothole by pothole, and other traffic-impacting features.
LVL5 applies machine learning to create its high definition (HD) maps, which differs from Waze and Apple Maps, the familiar 2D routing maps in common use on smartphones and newer vehicles. HD maps are machine-readable and 3D representations of static features (road signs, traffic lights, curbs, lane lines) in the driving environment. The degree of accuracy is between five and 50 centimeters, sufficient to detect meaningful pavement imperfections.
Can this help DOTs find and fix potholes more efficiently?
According to Andrew Kouri at LVL5, the technology is being evaluated by departments of transportation for road maintenance applications. While no municipality has yet to sign on to using it, “It can be a neutral voice,” suggests Kouri. “Cities can now know where those imperfections are. It’s impartial, not dependent on individuals who happen to be the type to phone in complaints.”
Already 11,000 drivers – who are paid by the mile to use the phone app on their dashboards – are participating in the Pavdar system and providing this data (drivers are paid between one and five cents per mile; many of the motorists are ride-share drivers). The company published a map in late 2018 that illustrated just how broadly it is in use. Just as important, this map ranked road quality by county, from best to worst. Across five million miles of US roads recorded by Pavdar, it found the best pavement is in Florida and the worst is in Michigan.
The company points out that Lidar-based systems lag in providing data to cities, where Pavdar is constantly updating because their collection methods are affordable and continuous. “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),” Kouri adds.
The firm also looked for correlations between high gas taxes and spending on construction in the effects of those things on road quality. Interestingly, there was no discernable correlation between higher spending and fewer potholes. It’s unclear if “spending on construction” encompassed only new building and repaving, or also preventive maintenance (crack sealing and timely repairs to small potholes with permanent-fix asphalt). And to be fair, Michigan undergoes far more freeze-thaw conditions than Florida, the leading cause of potholes (but not the only cause; moisture of all kinds, traffic and high temperatures also contribute to asphalt deterioration).
“State and federal governments spend over $400 billion on maintaining and building new roads each year,” says Kouri, raising important questions on infrastructure proposals, planning and execution.
