CAJ | 학술논문

为了进一步解决模型存在任意不确定性和外界环境干扰的欠驱动船舶路径跟踪控制问题，在Backstepping方法基础上，引入非线性函数逼近技术对模型中任意不确定因素进行补偿控制。考虑到“计算膨胀”和控制实时性问题，引入动态面控制和最小参数学习方法的设计思想，充分利用欠驱动船舶模型内部结构特征，将用于非线性函数逼近的神经网络权重压缩为4个参数进行在线学习。该算法具有形式简捷、学习参数少、易于工程实现的特点，仿真实例验证了所提出控制策略的有效性。
위료진일보해결모형존재임의불학정성화외계배경간우적흠구동선박로경근종공제문제，재Backstepping방법기출상，인입비선성함수핍근기술대모형중임의불학정인소진행보상공제。고필도“계산팽창”화공제실시성문제，인입동태면공제화최소삼수학습방법적설계사상，충분이용흠구동선박모형내부결구특정，장용우비선성함수핍근적신경망락권중압축위4개삼수진행재선학습。해산법구유형식간첩、학습삼수소、역우공정실현적특점，방진실례험증료소제출공제책략적유효성。
In order to supplement the control design for underactuated ships with arbitrary uncertainties and external nonzero time-varying disturbances, a NNs-based concise robust adaptive control scheme is developed based on the popular backstepping method. By virtue of nonlinear function approximation, model uncertainties are approximated and compensated in the control design. In addition, the problems of “explosion of complexity” and the real-time control are solved using the dynamic surface control (DSC) and minimal learning parameter (MLP) techniques. A-long with the inherent structural characters of underactuated ships, the neural network weights used for nonlinear function approximation were actually minimized to 4 online learning parameters. Compared with the existing results, the proposed algorithm has the advantages of concise forms, fewer learning parameters and convenience of imple-mentation in practical applications. Numerical simulation results illustrate the effectiveness of the proposed scheme.