Robots that coexist with humans in real-world environments must interact with their surroundings to perform various tasks. Robots acquire multimodal information about their environment through sensors such as depth sensors, RGB cameras, microphones, and wheel encoders. By processing this multimodal observation data, robots can learn environmental maps, recognize object positions, classify place categories, and estimate their state within the environment. Such spatial cognition abilities are essential for robots operating in real-world settings.