具有未知死区和增益符号的自适应神经网络控制

Adaptive neural network control with unknown dead-zone and gain sign

针对一类具有未知死区和未知控制增益符号的单输入单输出非线性系统,根据滑模控制原理,并利用Nussbaum函数的性质,提出两种自适应神经网络控制器的设计方案.通过引入示性函数,提出一种简化死区模型,取消了死区模型的倾斜度相等的条件.通过引入逼近误差的自适应补偿项来消除建模误差和参数估计误差的影响.理论分析证明闭环系统是半全局一致终结有界.仿真结果表明该方法的有效性.

The problem of adaptive neural network control for a class of single input single output （SISO） nonlinear systems with unknown dead-zone and function control gain sign is studied in this paper. Based on the principle of sliding mode control and the property of Nussbaum function, two design schemes of adaptive neural network controller are proposed. By introducing characteristic function for the dead-zone model in the systems, a simplified dead-zone model is developed. The approach removes the condition of the equal slope with defined region. The adaptive compensation term of the approximation error is adopted to minify the influence of modeling errors and parameter estimation errors. By theoretical analysis, the closed-loop control system is proved to be semi-globally uniformly ultimately bounded. Simulation results demonstrate the effectiveness of the approach.