An investigation into advanced predictive control methodologies for the REMUS AUV
University of New Brunswick
The health of the ocean ecosystem is vitally important to humans and the planet. In order to maintain the health of these environments, and reduce the human impact upon them, efforts must be made to monitor them. A direct method for monitoring ocean environments is the autonomous underwater vehicle (AUV). A well verified AUV model (the REMUS AUV) is chosen as the platform for this thesis. This vehicle model represents a highly nonlinear, underactuated, highly coupled system in three spatial dimensions. The three dimensional path-following of this system is addressed in this work. This thesis examines the effectiveness of various advanced predictive control algorithms in controlling simplified systems exhibiting similar properties to those listed above, to be extended to the AUV. The full nonlinear model of the AUV is run through various simulations and tuning is discussed. A simplified Line of Sight (LOS) waypoint-based guidance system is presented, along with an advanced virtual vehicle guidance system, controlled in two stages. The error model for the virtual vehicle is shown, along with the derivation of the first stage control method (constrained model-based predictive control). The second stage control is then developed using the simplified nonlinear model-based predictive controller. A unique hybrid guidance system is introduced, combining these two guidance methods. Simulations on the hybrid guidance system demonstrate the good path-following capability of both control systems and the guidance systems introduced.