Using external cameras to help robots map unfamiliar spaces faster

Robots exploring unknown environments now benefit from fixed external camera feeds—like security cameras or remote views—that create a geometric and semantic map before the robot moves. The system treats all camera feeds identically through a single RGB-based pipeline, fusing external views with onboard sensors to build 3D scene graphs incrementally. In experiments, even one external camera boosted initial object detection by 79% and made active exploration significantly more efficient.