Técnicas de localização para um veículo aquático baseadas em mapas batimétricos e filtro de partículas

• Autor(a) principal: França, Rodrigo Paz
• Data de Publicação: 2016
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_RS
• Texto Completo: http://tede2.pucrs.br/tede2/handle/tede/6676
• topic: SUBMARINOS; DETECTORES; FILTROS; SENSORES ELÉTRICOS; ENGENHARIA ELÉTRICA; ENGENHARIAS

This work presents localization methods for underwater vehicles based on sensor fusion. To achieve the sensor fusion two techniques were used: Extended Kalman Filter (EKF) and Particle Filter (PF). The EKF method developed uses an Inertial Measurement Unit (IMU) and a GPS Intelligent Buoys (GIB) sensor. With the purpose of developing a low-cost location method which does not require external sensors, the Terrain Based Localization (TBL) technique was also developed. This method consists in the fusion between an IMU, a sonar and a bathymetric map of the navigation region through the PF. This technique utilizes particles to estimate the probability function of the vehicle position, however, this approach shows poor precision in regions of low depth variation. In order to solve this problem two solutions are presented, one in software, which uses a trajectory generation algorithm (TG) that limits the vehicle navigation to regions of the map with large depth variation; and another solution which uses a peripheral GIB sensor, with the purpose of improving the PF correction step. In order to validate the developed methods simulations in the MATLAB software were made utilizing an AUV (Autonomous Underwater Vehicle) mathematical model. An analysis of the computational cost of the LBT technique was performed, through the implementation of the algorithm in the “C” programming language, where it was embedded on a micro-controlled board to measure the method execution time. In order to analyze the convergence capacity of the TBL and the TG algorithm effect thousand simulations with different quantities of particles were performed for TBL technique and TBL together with the GT algorithm.