神经自适应动态面控制与仿真

Adaptive Neural Network Control Using Dynamic Surface Control and Simulation

针对一类严格反馈不确定非线性系统,提出了一种神经自适应动态面控制器的设计方案。该方案改进了动态面的虚拟输入和真实输入以减小DSC的保守性,并将不确定分解为线性参数部分和非线性参数部分,两部分分别辨识。用一个BP神经网络逼近系统中所有的非线性参数部分,所需辨识的参数较少。通过合理选取设计常数,可保证所有信号最终一致有界且跟踪误差可收敛到原点的一个小邻域内。

A novel design scheme of adaptive neural network with dynamic surface controller was proposed for a class of strict-feedback nonlinear systems with unknown uncertainty.The approach utilizes improved virtual or actual input to reduce DSC conservative,at the same time,the uncertainty were separated into two parts：one is linearly parameters,the other is non-linear parameters which can be approached by a multi-output BP neural network so as to identify the parameters of less.The closed-loop systems is uniformly ultimately bounded,with tracking error converging to a small neighborhood of origin by appropriately choosing design constants.