Sviluppo di un sistema di localizzazione multimodale per un AGV nell'industria primaria dell'alluminio

The project, where this work takes place, aims to provide a proof of concept for the various elements composing the control scheme of an Autonomous Guided Vehicle(AGV) in an industrial scenario(aluminum production plant). In particular, in this thesis, the indoor and outdoor localization problems are considered, defining for each one a pose estimation technique fusing different sensors and localization methods. The outdoor pose estimation is based on the usage of Normal Distribution Transform (NDT) matching, wheel odometry, GNSS RTK devices (high accuracy global positioning systems) and Inertial Measurement Unit (IMU) readings. In the indoor localization method, almost the same techniques and sensors are used, substituting the GNSS readings with the estimates retrieved from a camera based algorithm. In order to combine the information retrieved by different algorithms and measuring devices, a fusion method, the Extended Kalman Filter (EKF), is used. Experimental tests have been carried out in the Gazebo virtual environment using the Robot Operative System 2 (ROS2): for each scenario, the simulated vehicle moved along a complex path to evaluate the localization reliability of the presented algorithms. Moreover, in the outdoor scenario, the localization scheme is also tested removing the satellites systems readings, in order to simulate a GNSS denied environment and prove the robustness is such conditions. This work has been realized thanks to the collaboration between the University of Padova and the Thecmo Car S.P.A. company, a world leader company specialized in the production of high-end mobile and stationary machine for the primary aluminum industry.