船舶航向保持的鲁棒神经网络控制  

Robust neural network control for ship's course-keeping

作　　者：张显库[1] 吕晓菲[1] 郭晨[1] 杨盐生[1] 

机构地区：[1]大连海事大学航海动态仿真和控制实验室,辽宁大连116026

出　　处：《船舶力学》2006年第5期54-58,共5页Journal of Ship Mechanics

基　　金：国家自然科学基金资助项目(60474014);国家博士点基金资助项目(20040151007);交通部基础研究资助项目(200432922504)

摘　　要：设计一个N5-5-5-5-5-5-1前向神经网络,经训练得到一个开环的船舶航向保持控制器,使其能够较好地保持典型输入信号作用下的船舶航向;然后应用闭环增益成形算法,形成闭环控制系统。这种基于神经网络的航向保持方案,将神经网络的适应能力、非线性函数映射能力和闭环增益成形算法的鲁棒性结合起来,改善了系统的鲁棒性能。这种方法设计过程简单,物理意义明显。A N5-5-5-5-5-5-1 feed-forward network is designed and trained to acquire an open loop controller of course-keeping for ships： The NNC controller can follow the tacks of typical course curve with good precision,then a closed loop control system is constructed using closed loop gain shaping algorithm. The robustness of the system is improved because the control strategy connects the adaptability and nonlinear mapping of neural network with robustness of closed loop gain shaping algorithm.This method has the advantages of simple design procedure and obvious physical sense.

关 键 词：神经网络 闭环增益成形 鲁棒控制器 航向保持 

分 类 号：U661.334[交通运输工程—船舶及航道工程]