Nonlinear Model Predictive Control and Adaptive State Estimation for Autonomous Wildlife Tracking in Marine Robotics

Marine wildlife observation is a component of marine studies that require precision and non intrusive autonomous platform. This thesis investigates the tracking of biological targets in dynamic marine environment using a BlueROV2 controlled by a Non Linear Model Predictive Control (NMPC). The research first adresses the challenges of the subsea domain, including unpredictable disturb maneuvering and hydrodynamic disturbances. To bridge the gap between noises, disturb and control, an Augmented Extended Kalman Filter for Disturbances (AEKFD) is implemented to provide robust state estimation and real time estimation of external disturbances. A "Shadow-leader" guidance law is developed to mantain a safe distance from the target, to also keep it withing the camera's field of view without causing behavorial distruption. The control system utilized the Acados solver for real time optimization of the vehicle's 6DoF trajectory. Simulation results demonstrate tracking stability and accuracy under environmental uncertainties