The work presented in this thesis focus on the localization and the navigation of the ATRV2 mobile system in an unknown environment. This mobile system must first model and build his map using data from its onboard sensors, localize itself in this map, and then, generate a collision-free path allowing to it to reach his goal. The model of the environment we have chosen is a metric model (certainty grid) based on the HIMM algorithm, and is updated using data from the onboard sensors. The used localization method is based on the extended Kalman filter (EKF). This method merges the odometric data (system model) and the data from the onboard perception system (measure) to regularly correct the relative estimate of the mobile system position due to the cumulative error caused by the odometer sensors. To determine the measure, the method uses the map matching while estimating its search area and the position around which it searches the estimated position. For the navigation, we have proposed a new method called DVFF combining path planning method based on the D* algorithm, and reactive non-stop method based on the virtual force field VFF. This navigation method uses the current values of onboard ultrasonic sensors (on a small time window), and data from the environment model, to decide what action to take. It allows to the ATRV2 mobile system to move without collisions with obstacles of its environment.
