论文部分内容阅读
基于神经网络所具有的灵活强大的学习能力,提出了一种用多层前馈神经网络实现的控制器.该控制器通过学习系统的逆动力学特性,能由系统反馈回的输入/输出状态及未来期望输出值直接得到应加在系统输入端的控制量.另外,通过引入系统的神经网络正向模型,可将系统输出端的误差经网络逐层反传,在线调节神经网络控制器的权重,从而使控制器具有自学习能力,以适应控制对象参数的变化,确保良好的控制效果.
Based on the flexible and powerful learning ability of neural network, a controller using multilayer feedforward neural network is proposed. By learning the inverse kinematics of the system, the controller can directly obtain the amount of control to be added to the system input by the feedback / input status and future expected output of the system. In addition, by introducing the forward neural network model of the system, the error of the output of the system can be inversely transmitted through the network layer by layer, and the weight of the neural network controller can be adjusted online so that the controller has the ability of self-learning to adapt to the change of the parameters of the controlled object , To ensure good control effect.