Startup Uses Machine Learning to Drive Better Road Investment
Emily Pollock posted on January 18, 2019 |
RoadBotics turns visual road scans into maps that show areas in need of repair. (Image courtesy of RoadBotics.)
Move over, RoboCop! The newest robot looking after Detroit’s streets is RoadBotics’ pavement-analyzing machine learning algorithm, which uses driver-collected data to determine which areas of the city are in most dire need of roadwork.
RoadBotics is a startup with a simple premise: you provide a map of the roads you want scanned, and it will scan them and tell you where the potholes are the biggest. To do that, the company has drivers travel down all the roads with dashboard-mounted smartphones recording a continuous stream of video and then cross-references that information with GPS data. After RoadBotics has a rough map, it analyzes the footage with a machine learning algorithm trained on common road defects (unsealed cracks, potholes, etc.) Finally, its platform aggregates the total damage on a section of road and assigns it a rating from 1-5 (with 1 being a practically new road, and 5 being a road that is in serious need of repairs).
For cities and states, the allure of RoadBotics is that it gives “objective” data without human analysis, and that it provides its users with an easy-interface map afterward. So far, the company has assessed over 90 communities in 15 states. Detroit is one of the most recent cities to be announced…