A Master of Science thesis in Mechanical Engineering by Ali H M Wadi entitled, "Modeling and Guidance of an Underactuated Autonomous Underwater Vehicle," submitted in December 2017. Thesis advisor is Dr. Jin-Hyuk Lee and thesis co-advisor is Dr. Shayok Mukhopadhyay. Soft and hard copy available.
Autonomous Underwater Vehicles (AUVs) have become an indispensable tool that is employed by an array of fields. From the inspection of underwater cables and pipelines, to the monitoring of fish pens and coral reefs, to the detection and disposal of mines, and to the executing search and rescue operations, AUV research and development has received a lot of attention. This thesis is concerned with the mathematical modeling of an underactuated AUV to execute its missions. The modeling task entails identification of the numerous parameters of the vehicle. A finite element analysis software was used to estimate the parameters describing drag and hydrodynamic mass phenomena. While the proposed underactuated configuration promotes the deployment of more energy-efficient vehicles, this configuration imposes complications on the guidance and motion control tasks as the vehicle becomes constrained in the way it can reach certain positions or perform certain motions (anholonomy). To tackle this trajectory tracking guidance problem, a model-based controller that overcomes the underactuated nature of the vehicle was designed. This controller was further enhanced by the novel development and application of a Universal Adaptive Stabilizer-based adaptation law that aims to minimize controller effort, reject noise, and provide robust trajectory tracking. The adaptation is governed by a statistical management system to ensure proper operation in a noisy underwater environment. Moreover, the navigation problem is touched upon by implementing a sensor fusion algorithm to estimate the vehicle state in its noisy environment. The algorithm investigates an Extended Kalman Filter as well as an Unscented Kalman Filter to fuse the available information from sensors with the modeled dynamics of the vehicle and provide better estimates of the vehicle state. Additionally, the hardware and software was integrated in a Robot Operating System setting, and a Gazebo-based simulation environment that enables the visual depiction and testing of algorithms on the considered AUV was developed. The parameter identification methodology compared well to published analytical and empirical forms, the proposed adaptation law outperformed traditional techniques like Adaptive Proportional Controllers, and the gain management system demonstrated excellent potential at maintaining stable operation of the vehicle in very noisy environments.