Abstract: An autonomous mobile device (AMD) uses sensors to explore a physical space and determine the locations of obstacles. Simultaneous localization and mapping (SLAM) techniques are used to process images from cameras to determine location and trajectory of the device. As the AMD moves and explores to generate an occupancy map, errors in the localization accrue. The occupancy map is processed to determine a graph that is based on, for a given point in physical space, a distance, relative direction, and a characteristic of a closest obstacle. A first graph based on a global occupancy map may be compared to a second graph based on a local occupancy map to determine a displacement between the local occupancy map and the global occupancy map. To facilitate loop closure, the AMD may navigate to a point along the first graph, where a loop closure process may be performed.