The demo presents the results of work on an autonomous system for mapping spaces and navigating within them using natural language to specify the target destination. The system consists of two components. One of them is an autonomous robot capable of moving indoors in a way that ensures comprehensive mapping of the space while meeting image acquisition requirements; that is, every area is captured in a photograph at the correct angle without excessive overlap between consecutive images. Each photograph of the space is tagged with metadata, including the coordinates where the shot was taken. During the navigation phase, the robot can move to a designated location and avoid obstacles.
The second component of the system is a module that processes images from the acquisition phase and stores all information extracted from the images in a database. In navigation mode, natural language queries are translated into a list of sought objects, which are then searched for in the database. Finally, potential locations are verified for the target, and the destination location is identified.
At IITiS, we are also developing indoor localization technology, a complementary solution to those presented in this demo.
Comments