The clues can be described in general terms, like “front door” or “garage.” As MIT explains, the robot might be trained to know that a driveway often leads to a sidewalk which likely leads to the front door.

“Even if a robot is delivering a package to an environment it’s never been to, there might be clues that will be the same as other places it’s seen,” said Michael Everett, a graduate student in MIT’s Department of Mechanical Engineering. “So the world may be laid out a little differently, but there’s probably some things in common.”

The technique uses pre-existing algorithms that looks at visual data and labels the scene with semantic clues, like “sidewalk.” The researchers used those algorithms to build a map as their robot moved around. What’s new about their approach is that they allowed the robot to make decisions and determine the most efficient path to a destination. The team will present their findings at the International Conference on Intelligent Robots and Systems this week.