This project joins within the framework of the societal challenge 6 " Mobility and sustainable urban systems " of the generic call for projects of the ANR 2017 in conformance with the instrument of financing Young Researcher ( CJC). It is interested in the improvement of the positioning for the navigation of the urban vehicles. This project is interested in the implementation of a geolocalized map of objects and semantic events adapted to the urban navigation. The peculiarity of these works lies in the implementation of a bidirectional strong interaction between the mapping and the semantic navigation in an innovative approach of simultaneous localization and semantic mapping (SLAM) using a tight fusion between a semantic visual approach and a GNSS positioning. Thus the objective is to study the possibilities offered by the active collaboration of a classical GNSS/IMU/Visual Odometry and of a SLAM approach based only on semantic objects for the improvement of the positioning in urban areas in the context of the autonomous navigation. This project so attacks two main aspects of an accurate navigation in urban areas: the precise positioning and the scene analysis and understanding. The ambition of this project is to integrate different levels of intelligence within the framework of the autonomous navigation, namely to integrate the semantic information into the low level navigation task (only based on the geometrical structure of the scene). For that purpose, surrounding area of Machine-Learning will be integrated to detect, extract and identify static and dynamic objects of the environment more easily recognizable on the long term or according to the weather or visibility conditions. A higher level of intelligence could also be studied concerning the analysis of the inter-object's interactions to include a certain semantic of action (accidents, pedestrians' crossing etc.). The long-term ambition is to create an intelligence of the situation, allowing a safe navigation in urban context.