CAJ | 학술논문

针对一类可转化为“标准块控制形”的MIMO非线性系统，基于动态面控制技术，提出一种鲁棒自适应神经网络控制算法。采用径向基函数神经网络逼近不确定性模型，通过引入一阶滤波器，消除后推设计中由于反复对虚拟控制的求导而导致的复杂性问题，同时补偿项的引入可避免反馈线性化方法中可能出现的控制器奇异性问题，无需控制增益矩阵正定、可逆的条件。利用李亚普诺夫方法，证明了闭环系统是半全局一致终结有界，适当选取设计常数，跟踪误差可收敛到原点的一个小邻域内。计算机仿真结果表明此法的有效性。
침대일류가전화위“표준괴공제형”적MIMO비선성계통，기우동태면공제기술，제출일충로봉자괄응신경망락공제산법。채용경향기함수신경망락핍근불학정성모형，통과인입일계려파기，소제후추설계중유우반복대허의공제적구도이도치적복잡성문제，동시보상항적인입가피면반궤선성화방법중가능출현적공제기기이성문제，무수공제증익구진정정、가역적조건。이용리아보낙부방법，증명료폐배계통시반전국일치종결유계，괄당선취설계상수，근종오차가수렴도원점적일개소린역내。계산궤방진결과표명차법적유효성。
Based on dynamic surface control, a systematic procedure for synthesis of robust adaptive neural network control is proposed for a class of MIMO nonlinear systems which could be turned to ＂standard block control type＂ ,with unneeded inverse gain matrix in this paper. By employing radial basis function neural net- works （RBFNNs） to approximate uncertain nonlinear system functions, the problem of explosion of complexity in traditional back stepping design, which is caused by repeated differentiations of certain nonlinear functions such as virtual control, is overcome by introducing the first order filter. Moreover, the possible controller sin- gularity in feedback linearization is avoided without projection algorithm. Using Lyapunov method, the closed- loop system is proved to be semi-globally uniformly ultimately bounded, with tracking error converging to a small neighborhood of origin by appropriately choosing design constants. Simulation results demonstrate the ef- fectiveness of the proposed method.