改进的神经网络观测器在非线性系统中的应用

为降低非线性观测器对模型精度的依赖性, 提出一种非传统的神经网络观测器设计方法。该神经网络为三层前馈网络, 采用带修正项的误差反传算法进行训练, 以保证控制的精度和权值有界, 利用神经网络识别系统的非线性部分, 并结合传统的龙伯格观测器重构系统状态; 利用Lyapunov 直接法保证基于权值误差的非观测器的稳定性, 并将该观测器应用于机器人轨迹跟踪控制中。仿真结果表明, 该方法解决了模型不确定系统状态观测问题, 适用于模型精度较低的非线性系统。

Improved Neural Network State Observer Designed for Nonlinear System

For reducing the dependence of nonlinear observer on the precision model, a non conventional NN(Neural Network) observer for nonlinear system is proposed. The neuro-observer is a three-layer feedforward neural network, which is trained extensively with the error backpropagation learning algorithm including a correction term to guarantee good tracking and bounded NN weights. Designing the neural network observer is using artificial neural network to identify the nonlinear parts of the system and using a Luenberger observer to reconstruct the states of the system. The Lyapunov direct method is used in order to ensure the stability of the proposed non-conventional observer. The proposed observer is applied to 2 degrees of freedom horizontal manipulator to evaluate its performance. The simulation results show that the state observation of uncertain systems can be solved by the method and it is suitable for the low precision model of the nonlinear system.