Core Insights - The Ministry of Natural Resources' First Institute of Oceanography has made significant advancements in unmanned vessel modeling and control, providing new technological pathways for autonomous navigation and intelligent control of unmanned vessels [1] Group 1: Challenges in Current Technology - The current technology faces two main challenges: the difficulty in precise modeling of high-speed unmanned vessels under multiple navigation states, and the struggle to balance reliability and interpretability in control methods [1] - High-speed unmanned vessels exhibit strong non-linear and multi-stage characteristics, making traditional physical models inadequate [1] - Existing deterministic model-based control methods lack robustness, while current neural network controllers often lack physical modeling, leading to insufficient interpretability [1] Group 2: Research Developments - The research team conducted a series of navigation experiments using the self-developed 7-meter high-speed unmanned vessel "Jiuhang 750" to address the issues of physical-data modeling and learning control [2] - A hybrid modeling and control framework based on the Koopman physical model and neural networks was proposed, which offers advantages in interpretability and accuracy, enabling reliable control of unmanned vessels in complex marine environments [2] - The proposed control framework was validated through comparative experiments, demonstrating superiority over traditional nonlinear control methods, thus providing new technical support for the safe autonomous navigation of unmanned vessels [2] Group 3: Modeling Techniques - To address the precise modeling of unmanned vessels at low, medium, and high speeds, the research team introduced a Gaussian process regression modeling method based on mixed kernel functions [2] - This method allows the unmanned vessel model to capture smooth global trends while also describing local abrupt changes, establishing a non-parametric motion model suitable for different navigation states [2] - Experimental validation showed that this model's accuracy surpasses traditional methods, providing effective means for multi-state unmanned vessel motion modeling and forecasting [2] Group 4: Implications for Industry - The series of research outcomes provides new modeling tools and control strategies for autonomous navigation and intelligent control of unmanned vessels in complex marine environments [2] - These advancements are expected to further promote the application of unmanned vessel systems in marine observation and operations [2]