The project focuses on the field of map construction and localization using omnidirectional vision, advancing towards a hybrid topological-metric paradigm, which allows (a) the incremental construction of a semantic map as the robot explores the unknown environment and (b) the estimation of the robot's position and orientation with precision, with 6 degrees of freedom and at a reasonable computational cost. Additionally, to improve the integration of the mobile robot in real social environments, where it must interact with people, some features will be included in the model to make it compatible with human perception. Thus, the proposal aims to go beyond the concept of multi-level hierarchical localization, adapting it to extensive and complex social environments, and including collaboration with users through high-level commands. This proposal is organized around two main lines of research:

• Line A: Incremental creation of hybrid metric-topological maps from the global appearance of a set of scenes.
• Line B: Construction of environment models that allow localization with 6 degrees of freedom from visual information.