Neuro adaptive guidance and control system of a multipurpose autonomous underwater vehicle for underwater environment monitoring

This thesis introduces a novel neuro-adaptive control framework for Blucy, an hybrid unmanned underwater vehicle (UUV), designed for environmental monitoring in complex underwater environments. Central to this work is the development of a comprehensive benchmark model, rigorously validated against real mission data, which serves as a foundational simulator for designing the neuro-adaptive controller. An original parameter identification workflow is proposed, integrating CAD modeling, CFD simulations, and AMCOMP software to accurately capture hydrodynamic forces and thruster dynamics critical for precise modeling. Based on the benchmark, a novel fixed-time sliding mode control system is designed, augmented with neural networks and disturbance observers that estimate uncertainties and disturbances. To enhance adaptability, the neural network and disturbance observer are trained using a composite error learning strategy, optimizing real-time response to environmental changes and system faults. The effectiveness of the neuro adaptive control framework ensuring precise trajectory tracking and robustness against uncertainties, disturbance and faults, is demonstrated through extensive simulations.