This paper addresses the issue of waypoints-based path-following control for sailboats. Based on the parameterized line-of-sight framework, a novel guidance principle for the sailboat is developed, which comprises the mechanisms to observe the crab angle and generate the reference heading angle. For the crab angle observation, the approach with double reduced-order extended state observers and a saturation operator is developed, which involves the consideration of the intense sway. For the heading angle generation, three modes are devised, i.e., the path-following mode, the characteristic tacking mode, and the gybing mode. Sign functions are employed to describe tacking and gybing maneuvers of the sailboat such that rational switching of the reference heading angle is guaranteed, and no linguistic command is required. Matching with the proposed guidance principle, an adaptive neural control law is thereafter developed with the outputs landing in rudder angles. The echo state networks (ESNs) are introduced to offset model uncertainties and environmental disturbances with minimum parameters learning. By introducing two time-varying systems, all the errors in the closed-loop system are proved to be semiglobal uniformly ultimate bounded. Finally, the effectiveness of the proposed scheme is testified in MATLAB simulation.

中文翻译：

---

帆船的基于视线的制导和自适应神经路径跟随控制

本文解决了帆船的基于航点的路径跟随控制问题。基于参数化视线框架，开发了一种新的帆船制导原理，包括观察偏航角和生成参考航向角的机制。对于偏斜观测，开发了双降阶扩展状态观测器和饱和算子的方法，这涉及到强烈摇摆的考虑。对于航向角生成，设计了三种模式，即路径跟随模式、特征跟踪模式和转向模式。使用符号函数来描述帆船的跟踪和转向操作，从而保证参考航向角的合理切换，并且不需要语言命令。与所提出的制导原理相匹配，此后开发了一种自适应神经控制律，输出以舵角着陆。引入回声状态网络 (ESN) 以通过最小参数学习来抵消模型不确定性和环境干扰。通过引入两个时变系统，证明闭环系统中的所有误差都是半全局一致极限有界的。最后通过MATLAB仿真验证了所提方案的有效性。证明闭环系统中的所有误差都是半全局一致极限有界的。最后通过MATLAB仿真验证了所提方案的有效性。证明闭环系统中的所有误差都是半全局一致极限有界的。最后通过MATLAB仿真验证了所提方案的有效性。