| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 37.31 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A necessidade de exploração e mineração no meio subaquático desencadeou, de
alguma forma, uma acentuada evolução nas tecnologias e no desenvolvimento de
veículos autónomos que opere nesse tipo de ambientes. A autonomia nas operações
encontra-se, inteiramente, relacionada com o sistema de localização e é nesse sentido
que o desenvolvimento desta dissertação se foca.
Os resultados obtidos com o presente estudo pretendem constituir um contributo
para a obtenção de maior precisão e exatidão na performance de um sistema de
localização subaquático. Neste sentido, são estudados dois aspetos fundamentais
para um sistema de localização ser robusto e preciso, sendo eles a fusão sensorial e
a calibração.
A fusão sensorial consistiu em utilizar um EKF com dados sensoriais do DVL, USBL,
acelerómetros e sensor de pressão para estimar a localização. No entanto, para a
realização da fusão sensorial, os sensores necessitam de ser calibrados, pois o desalinhamento
entre os diversos sensores produz influência negativamente no sistema de
localização. Assim sendo, foi desenvolvido um sistema de calibração para minimizar
esse problema.
As Minas subaquáticas, é um dos principais locais de operação para UUV para o
INESC TEC, no entanto, as Minas apresentam distorções magnéticas sendo uma
grande dificuldade para os sistemas de localização tradicionais. Deste modo, foi
proposto um modelo para a obtenção da atitude sem recurso aos magnetómetros
evitando a essas distorções.
Por fim, com o intuito de efetuar a validação dos algoritmos desenvolvidos foram realizados
datasets para comprovar a eficiência do sistema de localização subaquático
proposto.
Os resultados obtidos com o sistema desenvolvido neste estudo, quando comparados
com o ground truth apresentaram erros inferiores a meio metro na posição estimada
e 0.1 graus na orientação estimada, ao longo de um percurso de 722 m com uma
duração de 10 minutos. É de salientar que o filtro probabilístico demonstrou-se
robusto mesmo com a existência de algumas falhas na obtenção das medidas dos
sensores acústicos.
The need to explore and mine underwater has triggered, in some ways, a pronounced evolution of the technologies and the development of autonomous vehicles which could operate in that type of environment. The operations autonomy is entirely related with the localization system and this thesis is focused in that sense. The results we obtained with the present study aim to contribute to the process of obtaining a higher precision and accuracy in the performance of an underwater localization system. In this sense, we study two fundamental features for the robustness and precision of a localization system: multiple sensor fusion and system calibration. An Extended Kalman Filter was used to fuse the sensor data from Doppler Velocity Log, Ultra-short baseline, accelerometers and a pressure sensor to estimate the position. However, in order to perform the sensorial fusion, the sensors need to be calibrated since the misalignment between the different sensors produces negative effects on the localization system. Thus, we have developed a calibration system to decimate this problem. The underwater mines, one of the means of operation for the autonomous vehicles used in some ongoing projects at the INESC TEC's Centre for Robotics and Autonomous Systems, present very hostile characteristics with high magnetic distortions. This being the case, we have proposed a model to obtain the attitude without resorting to the magnetometers which would be immune to these distortions. Lastly, in order to validate the algorithms we have developed, we have performed datasets to prove the eficiency of the underwater localization system we propose. The results we obtained with the system we have developed in this study, when compared to the ground truth, presented errors which were inferior to half a meter in the estimated position and 0.1 degrees in the estimated orientation, through a route of 722 meters with a duration of 10 minutes. We should emphasize that the probabilistic filter proved to be robust even with the existence of some aws while obtaining the measurements of the acoustic sensors.
The need to explore and mine underwater has triggered, in some ways, a pronounced evolution of the technologies and the development of autonomous vehicles which could operate in that type of environment. The operations autonomy is entirely related with the localization system and this thesis is focused in that sense. The results we obtained with the present study aim to contribute to the process of obtaining a higher precision and accuracy in the performance of an underwater localization system. In this sense, we study two fundamental features for the robustness and precision of a localization system: multiple sensor fusion and system calibration. An Extended Kalman Filter was used to fuse the sensor data from Doppler Velocity Log, Ultra-short baseline, accelerometers and a pressure sensor to estimate the position. However, in order to perform the sensorial fusion, the sensors need to be calibrated since the misalignment between the different sensors produces negative effects on the localization system. Thus, we have developed a calibration system to decimate this problem. The underwater mines, one of the means of operation for the autonomous vehicles used in some ongoing projects at the INESC TEC's Centre for Robotics and Autonomous Systems, present very hostile characteristics with high magnetic distortions. This being the case, we have proposed a model to obtain the attitude without resorting to the magnetometers which would be immune to these distortions. Lastly, in order to validate the algorithms we have developed, we have performed datasets to prove the eficiency of the underwater localization system we propose. The results we obtained with the system we have developed in this study, when compared to the ground truth, presented errors which were inferior to half a meter in the estimated position and 0.1 degrees in the estimated orientation, through a route of 722 meters with a duration of 10 minutes. We should emphasize that the probabilistic filter proved to be robust even with the existence of some aws while obtaining the measurements of the acoustic sensors.
Description
Keywords
Localização subaquática USBL DVL Calibração DVL Calibração USBL Fusão sensorial AHRS EKF Underwater Localization Calibration DVL Calibration USBL Data Fusion