Unmanned Underwater navigation based on visual and inertial sensors

Unmanned Underwater Vehicles (UUVs) are widely used for decades mainly in exploration missions and underwater maintenance. Autonomous Underwater Vehicles (AUVs) represent a class of this vehicles that is becoming more and more popular but they have high costs due to the price of the sensors needed for autonomous navigation purpose. This thesis has been focused on the estimation of UUV trajectories, combining Visual Odometry (VO) and Inertial Measurement Unit (IMU) approach. The challenge was to apply in an underwater environment methods commonly used in submerged scenarios. Another important aspect of this study was the utilization of low-cost hardware for data acquisition and manipulation. All algorithms were designed to run in a Raspberry Pi, a single camera for VO and an entry level IMU were used. A characterization of IMU was done in order to prevent from gyroscope’s drifts over the time. The fusion of VO and IMU measurements was done using an Extended Kalman Filter (EKF). Testing phase was divided in three steps: first tests were centered on performance of feature based VO algorithms in an underwater dataset, improving performance of these using image enhancement techniques. The second phase was focused on VO algorithm with the low-cost hardware, working in a submerged environment. The last phase consisted in a series of underwater tests of the complete algorithm.