混沌光学系统的神经网络自适应控制研究

Adaptive Control of Chaotic Optical Systems with Neural Networks

提出一种用于混沌光学系统控制的神经网络自适应控制技术。以一前向神经网络作为受控混沌光学系统的系统辩识器，由此神经网络系统辩识器与受控混沌光学系统输出差值作为负反馈对受控混沌光学系统控制参数进行调整达到控制目的。由于所使用神经网络系统辩识器在常规ＢＰ算法的支持下可从受控混沌光学系统的输出时间序列进行动力学模型重构，因而特别适用于对未知动力学表述的混沌光学系统进行控制。以对布喇格声光双稳混沌系统的系统辩识及自适应控制为例，对此神经网络自适应控制技术可行性进行了示例证明。

An adaptive algorithm for controlling the chaotic optical system, termed as the neural network adaptive control algorithm in which the system identifier of the controlled chaotic optical system is a BP NN, is presented in this paper. In this NN adaptive algorithm the difference between the output of the controlled chaotic optical system and that of the trained NN identifier is employed as the negative feedback in readjusting the system's control parameter. Because the NN identifier can reconstruct the dynamics of the controlled chaotic optical system from the output time series with the support of the BP algorithm, this adaptive control algorithm is considered applied to control the chaotic optical system when the reference model for control can not be obtained in advance. The feasibility of this NN adaptive control algorithm is demonstrated with the computer simulation of identifying and then controlling a Bragg diffraction acoustooptic bistable system.