AI is turning connected cars into pothole-finding machines | TechCrunch

"Samsara has spent the last decade giving its customers cameras to mount inside millions of trucks for driver monitoring, theft prevention, and helping with liability claims. The San Francisco-based company has taken all that data and trained its own model that can detect multiple different types of potholes and determine how quickly they are deteriorating."

"The idea is that Samsara-equipped trucks are far more prevalent than Waymo's robotaxi fleet, which currently stands at just around 3,000 vehicles. Even as that number grows, Samsara believes it will be able to collect more data and, crucially, more repeat data from the same locations that show how potholes change over time."

"Samsara believes this data will be valuable to cities - the company announced Tuesday that the city of Chicago is already under contract as a customer - and that it will be the first in a series of insights and data points that will be offered in Ground Intelligence. Other potential features include detecting graffiti, broken guardrails, low-hanging power lines, or really "anything that we can observe that has relevance to a city, or also to the private sector," said Samsara's vice president of product, Johan Land."

"Typically, Land said, cities have to either dispatch workers or sift through hundreds of 311 calls to find these"

Potholes remain a persistent municipal problem despite prior claims that technology could mitigate them. As vehicles gain advanced sensors, they can quickly alert cities to potholes and other infrastructure issues. Waymo and Waze have piloted sharing pothole data with local governments, and Samsara offers an AI-powered system called “Ground Intelligence.” Samsara uses data from cameras mounted inside millions of trucks to detect multiple pothole types and estimate how quickly they are deteriorating. Because truck fleets are more widespread than robotaxi fleets, Samsara expects more and repeat observations from the same locations. The resulting data is intended to provide actionable insights for cities, including Chicago, and may extend to detecting graffiti, broken guardrails, low-hanging power lines, and other observable hazards relevant to public and private sectors.

Read at TechCrunch