The wheeled robot, developed in Prof. Angela Schoellig’s Learning Systems and Robotics Lab, uses a camera to generate a constantly updated three-dimensional map of a room while software identifies objects and their relevance to human activity.

For example, when searching for misplaced glasses in a kitchen, the robot understands that items are more likely to be found on tables or window sills than on a stovetop or in a sink. Using this contextual knowledge, the system calculates probabilities across the room and prioritizes the most likely locations.

The system also draws on internet-based knowledge to understand how people typically use objects in a space. Depth information from camera images allows the robot to assign probability values to likely search locations.

Researchers say the approach improves search efficiency by nearly 30 percent compared with random exploration. The robot also compares past and current images of its environment, allowing it to detect environmental changes with about 95 percent accuracy.

Future versions could search inside drawers or cupboards, which would require robots capable of opening doors and interacting with objects in their environment. You may never have to worry about finding your keys again!