In this paper, a robust adaptive neural networks control based on minimal–parameter-learning (MLP) is proposed for dynamic positioning (DP) of ships with unknown saturation, time delay, external disturbance and dynamic uncertainties. Through the velocities backstepping method, radial basis function (RBF) neural networks and robust adaptive control are incorporated to design a novel controller of which an appropriate Lyapunov-Krasovskii Function (LKF) is constructed to overcome the effect caused by time-delay. Meanwhile, the MLP technology is applied to reduce the computational burden while only one parameter need to be update by an adaptive law. In additional, a robust adaptive compensate term is introduced to estimate the bound of the lumped disturbance including the unknown saturation, unknown external disturbance and the approximate error of neural networks control while the robustness of MLP is improved and the unknown saturation is compensated. The developed control law makes the DP closed-loop system be uniformly ultimately stable which can be proved strictly through Lyapunov theory. Finally, simulations with a guidance law are proposed to demonstrate the validity of controller we developed.

本文针对饱和度，时滞，外部干扰和动态不确定性未知的船舶，提出了一种基于最小参数学习（MLP）的鲁棒自适应神经网络控制方法。通过速度反推法，结合了径向基函数神经网络和鲁棒自适应控制，设计了一种新颖的控制器，该控制器构造了适当的李雅普诺夫-卡拉索夫斯基函数（LKF），以克服时间延迟带来的影响。同时，MLP技术被应用来减少计算负担，而仅一个参数需要通过自适应定律来更新。此外，引入了鲁棒的自适应补偿项以估算包括未知饱和度在内的集总干扰的范围，未知外部干扰和神经网络控制的近似误差，同时提高了MLP的鲁棒性并补偿了未知饱和度。发达的控制律使DP闭环系统最终最终保持一致稳定，这可以通过Lyapunov理论严格证明。最后，提出了具有指导律的仿真，以证明我们开发的控制器的有效性。