A simple configuration for approximate learning models

The application of simple-pole configurations as learning models in self-optimizing control systems is considered, in particular for the case when the model must be an approximate plant representation. Theoretical bases are presented for evaluating a model's adequacy as a simulator and predictor within a control system; and it is shown that a model with a variable multiple time constant and variable gain will often be the best simple configuration. This type of model is likely to be useful as a self-adjusting learning model because it has only two parameters, each of which has a significant effect on the response.