Unmanned Surface Vehicles (USVs) play a pivotal role in various applications, including surface rescue, commercial transactions, scientific exploration, water rescue, and military operations. The effective control of high-speed unmanned surface boats stands as a critical aspect within the overall USV system, particularly in challenging environments marked by complex surface obstacles and dynamic conditions, such as time-varying surges, non-directional forces, and unpredictable winds. In this paper, we propose a data-driven control method based on Koopman theory. This involves constructing a high-dimensional linear model by mapping a low-dimensional nonlinear model to a higher-dimensional linear space through data identification. The observable USVs dynamical system is dynamically reconstructed using online error learning. To enhance tracking control accuracy, we utilize a Constructive Lyapunov Function (CLF)-Control Barrier Function (CBF)-Quadratic Programming (QP) approach to regulate the high-dimensional linear dynamical system obtained through identification. This approach facilitates error compensation, thereby achieving more precise tracking control.

翻译：无人水面艇（USVs）在诸多领域发挥着关键作用，包括水面救援、商业交易、科学探索、水上救援及军事行动。高速无人艇的有效控制是整体USV系统中的关键环节，尤其在复杂水面障碍物与动态条件并存（如时变涌浪、无定向力及不可预测风速）等具有挑战性环境中。本文提出一种基于库普曼理论的数据驱动控制方法：通过数据辨识将低维非线性模型映射至高维线性空间，构建高维线性模型；利用在线误差学习动态重构可观测USV动力学系统。为提升跟踪控制精度，我们采用构造性李雅普诺夫函数（CLF）-控制障碍函数（CBF）-二次规划（QP）方法，对辨识所得高维线性动力系统进行调节。该方法可实现误差补偿，从而获得更精确的跟踪控制。