一类非线性离散时间系统的变结构控制

Variable Structure Control for Nonlinear Discrete-time System

针对一类非线性离散时间系统,提出了一种基于神经网络趋近律的变结构控制方法。分别用两个神经网络自适应调整趋近律中的参数ε和δ,利用离散趋近控制律与离散等效控制律的偏差对网络权值进行更新,克服了常规变结构控制方法中需预先设定趋近律中参数的限制,既保留了传统趋近律设计方法的所有优点,又有效的改善了系统的动态品质,消除了系统抖振,使系统最终以理想方式在滑模面上运动。理论分析和仿真结果表明了所提出方法的有效性。

A variable structure control method based on neural network reaching law for a class of nonlinear discrete-time systems was proposed. Parameters, ε and δ, which were determined previously in the conventional reaching law, were regulated adaptively by two neural networks respectively, and network weights were updated by the deviations between discrete reaching control law and discrete equivalent control law. It is shown that all advantages of the reaching law are retained, meanwhile the dynamic features of the control system are improved effectively and system chattering is eliminated. System can move perfectly on the sliding-mode surface. Theoretic analysis and simulation results prove the validity of the method.