非线性系统的神经网络学习控制

Learning Control of Nonlinear Systems Using Neural Networks

本文将多层前向传递神经网络应用于非线性系统控制。通过对神经网络的训练,实现了非线性系统的状态反馈控制。这种方法不完全依赖于非线性对象的数学模型,因而为非线性系统的学习控制提供了一种有效的研究方法。用该方法对倒立摆进行实时控制,得到了满意的结果。

Multi-layered feed forward neural networks are applied to nonlinear learning control systems. By training the neural networks, optimal state feedback control of the non-linear system can be realized. This novel learning control mechanism, which does not depend upon too much about the nonlinear system, gives an exciting alternative for studying nonlinear control system. Real-time control to inverted pendulum shows that the new scheme is efficient to a large unknown nonlinearity.