Executive Summary : | An autonomous underwater vehicle, or unmanned underwater vehicle, can perform tasks with minimal human intervention. It consists of sensors, a computational unit, and actuators. These autonomous vehicles are used in various applications, such as surveillance, oceanographic data sampling, mine countermeasures, and bathymetry mapping. To track trajectories, path following, or way-point tracking, waypoint tracking algorithms will be implemented. To develop a guidance algorithm, a vehicle model is needed, which can be obtained through tow-tank tests or CFD analysis. However, these experiments are time-consuming and expensive. Instead, neural networks will be used to identify unknown parameters and design a guidance controller based on the identified model. The proposed adaptive guidance algorithm will incorporate actuator constraints and state constraints. Once the vehicle is developed and an adaptive guidance algorithm is designed, it will be extended for complete coverage of a given region. For example, a list of desired waypoints must be generated offline for complete coverage of a region. Other criteria and constraints include choosing a path that minimizes overall distance or performing complete coverage with minimum turnings. The project aims to develop a portable autonomous underwater vehicle, a guidance algorithm based on reinforcement learning, and a path planning algorithm with complete coverage. These algorithms will be verified in the field and simulation. |