Image Goal Navigation

Image goal navigation focuses on enabling robots to navigate to a location depicted in a target image, relying solely on onboard visual input. Current research emphasizes efficient map representations (e.g., lightweight topological maps, 3D Gaussian splatting), robust object recognition across viewpoints (leveraging contrastive learning and multi-view training), and effective exploration strategies (incorporating working memory and exploration-verification-exploitation frameworks). These advancements are crucial for developing more capable and adaptable robots for applications in assistive robotics, service robotics, and other real-world scenarios requiring autonomous navigation in complex, unstructured environments.