The Anticipatory Aspect in Neural Network Control

Abstract

The anticipatory aspect, essentially characteristic of high level intelligence systems, introduced in control systems by Neural Network controllers, gives an “intelligence” to the control system, as it becomes able to “remember” its past behaviour and use it afterwards in control algorithm in order to establish the appropriate control signal. Anticipation, in Neural Network control, is based in prediction capabilities of Multi Layer Perceptron, used in plant model construction, essential element of control process. In this work, the anticipatory aspect in Neural Network, different kind, controller’s structure is examined, and comparison between classic and Neural Network control is made. Ill. 5, bibl. 11 (in English; summaries in Lithuanian, English and Russian).