Navegação odométrica assistida por sistema de posicionamento global para agricultura de precisão

Abstract

Nowadays, agricultural vehicle navigation is almost exclusively dominated by global navigation satellite systems (GNSS), in particular, the global positioning system (GPS). However, when there are submetric position constraints, typical of the so-called "precision agriculture", this technology is not financially accessible to small and medium farmers, who correspond to more than 80% of farmers in Brazil (family farming). The main cause of this inaccessibility lies in the high cost of the employed GNSS devices, which are based on advanced differential and relative techniques. Despite GNSS, vehicular navigation can also be achieved through other technologies, such as inertial navigation systems (INS), cameras, radars, odometers, etc. It is known that odometric navigation, in particular, provides high precision in the short-term, but it presents cumulative errors that rapidly increase as the vehicle moves. This hinders the use of only odometric navigation for terrestrial vehicles subject to submetric positioning constraints. Despite this, odometry can serve as a base (and complementary) navigation technology to GNSS, which, unlike the former, provides good positioning accuracy in the long-term. This work aims at proposing, simulating, testing and evaluating the sensor fusion/integration of odometric navigation with GPS, for the purpose of submetric positioning of land vehicles, especially agricultural ones. As the main contribution of this work, two of the main odometry/GPS integration techniques are compared, namely, the loosely coupled and tightly coupled topologies, both implemented in the framework of an extended Kalman filter (EKF). Simulated and experimental results, conceived in MATLAB® software, are presented and analyzed, indicating the functionality, effectiveness and robustness of the proposed integrated navigation topologies. As an additional contribution of the work, an evaluation of the degradation of the performance of the investigated navigation solutions is proposed as function of the use of odometry in the tractioned and not tractioned wheels of the vehicle. In general, the loosely coupled integration performs better for the coupling on non-tractioned wheels, whereas the tightly coupled integration, for the coupling on tractioned wheels.