Topological Semantic Graph Memory for Image-Goal Navigation

Navigation with rich visual observations has been a critical issue in a variety of embodied agent tasks, such as exploration, image goal navigation, and object goal navigation [1-16]. A crucial ingredient for successful visual navigation is to construct a memory, which can represent the structure of the environment along with compact visual features for representing high-dimensional visual inputs. A metric-map memory [5, 17] created with SLAM, and a graph memory [8, 9, 18-20] with nodes and edges are the two standard memory construction approaches for navigation algorithms. Even though navigation systems that use metric maps produce powerful results with exact localization and mapping, it is not practical because the navigation agent is susceptible to sensory noises. The topological map, which represents geometric properties and spatial relations of places in the form of a graph, is proposed to construct a map without accurate mapping.