Landslides are difficult to predict, but a Carnegie Mellon University robotics researcher is working to create an early-warning system using “deep learning.” This type of artificial intelligence programs computers to recognize patterns and then make predictions based on existing data.
CMU’s Christoph Mertz uses photographs of hillsides around Pittsburgh, which computer algorithms analyze to identify and calculate where a landslide is more likely to occur.
“You detect those things and then you can do statistics on them,” said Mertz. “Like you know, if you see that there is more dirt than there used to be. Or that the crack has become more larger, or more frequent, or has changed very recently.”
Mertz predicts this project could be completed in about five years, when it may be even more crucial than it is today.