非线性系统的多神经网络自学习控制
NONLINEAR AUTO-LEARNING CONTROL USING MULTI-NEURAL NETWORKS
-
摘要: 本文提出了一种未知非线性动力学系统的多网络自学习控制方法．通过对系统的神经元网络辨识器和神经元网络控制器的有机结合，发展了基于逆动力学辨识器的控制网络广义Delta学习规则，从而使得整个控制系统具有很强的自适应、自学习能力．文中最后通过对系统进行的仿真研究证实了这种控制结构的有效性，仿真例子说明经过100个周期学习后，其系统的跟踪误差控制在1%以内．Abstract: This paper proposes a multi-network auto-learning control structure of the unknown nonlinear dynamic system. The generalized Delta rule of the controller network has been developed based on the identifier of the inverse dynamics.The controller architecture is presented with simulations demonstrating itsadaptive and learning abilities.It is also shown in the simulation that the maximum tracking error is within 1% after 100 periods.
下载:
