Fusão de sensores para localização: desenvolvimento de modelos para um veículo robótico subaquático

Abstract

Autonomous and teleoperated mobile vehicles use grows in several areas, aiming different activities for localization, and estimate position and orientation are important data that allow the robot to identify and perform it more accurately. In an underwater environment, using sensors, such as Inertial Measurement Unit, Inertial Navigation System, and wheels odometry, are essential when the environment does not allow the use of global data. In this work, sensors and techniques for localization were studied and simulated, using a ROS robotics framework and a 3D Gazebo simulator. After that, experiments with the Husky vehicle were performed, varying parameters, movement, and environment characteristics that allowed to evaluate the efficiency of the sensor fusion algorithms. In addition, an INS (Inertial Navigation System) model was developed for testing in the MCCR vehicle (MEC Combi Crawler Robot), a robotic vehicle for ship hull inspection, simulated in the Gazebo in a test tank, as well as a model for converting sonar data to PointCloud2. Subsequently, these models were used with sensor fusion algorithms (particles filters, extended and unscented Kalman filter) for error reduction in the simulated MCCR robot. The results highlighted the importance of properly configuring the fusion model parameters and the variability of the slippage impact on the odometry data. Therefore, an effective optimization of the fusion parameters for the MCCR robot will be possible after the odometry system characterization and validation of its model.