多变量过程自组织学习控制新方法及其实时实现

本文论述一种利用相联存储器系统的多变量自组织学习控制新方法及其实时实现。在线学习包含两个步骤:被控过程的动态输入—输出特性用仿视感控器的方式逐步存入一个相联存储器中,用来作为预测过程模型;依靠一种特殊的输出预测算法,利用预测过程模型并通过单步超前性能指标最优化来规划控制策略,经教练后的控制策略以同样的方式存入第二个相联存储器中。文中提出了一种特殊的采样和更新方法。通过仿真和实时应用证明了所述学习控制新方法的可行性。

A NEW LEARNIG CONTROL METHOD FOR MULTIVARIABLE PROC ESSES AND ITS REAL-TIME IMPLE MENTATION

The paper discuses a new self-organizing learning,multivarlable controlmethod utilizing associative memory systems and its real-time implementation.Theon-line learning compriss two stages:the dynamic input-output behaviour of the processto be controlled is stored stepwise in a perceptron-like manner into an associative memo-ry as a predictive process model,it uses then a special output predictive algorithem bywhich the control strategy planning via optimization of a 1-step ahead performance indexusing the predictive process model for performance mdex estimation is trained in the sameway into a second associative memory.For the read-time implementation of theaigorithem a special sampling and updating scheme is developed The efficiency of the newlearning control method is demonstrated by simulation and on-line application.