| dc.description.abstract | The usages of Remotely Operated Unmanned Underwater Vehicles (ROVs') for inspection of marine structures is today an essential part of the offshore industry. However, in the case of inspection of fish farms, the ROV has to spend continuous time relatively close to the water surface, in exposed seas, while navigating irregular flexible fish nets. The ability for an ROV to effectively navigate within the fish farm cage and guarantee a complete inspection requires new tools, and dynamic positioning (DP) improvements. This thesis will use the simulation tool FHSim to investigate two aspects regarding these issues. First, it will examine, by simulations, the performance of different variations of a nonlinear observer and an extended kalman filter (EKF) with or without a cascaded current observer, in the ROV's operating conditions. Secondly a visualization tool is integrated with the ROV DP system that allows it to traverse the fish net autonomously while also visualizing its position in the cage. The visualization module also identifies net areas that have been inspected by the ROV. All the modified DP modules are assessed in several simulation cases designed to mimic ROV inspection behaviour. It was found that in particular, the nonlinear observer performed better in estimation and wave filtering abilities, while the EKF was superior in dead reckoning. These differences were attributed to the different nature of the two observers, their programmed integration method, and possibly insufficient tuning of the EKF. The autonomous net traversal algorithm was successfully able to make the ROV traverse the cage, however, some trajectory faults were identified. Improvements to overcome these faults were suggested. | |
