Accession Number : AD0624548

Title :   AUTOMATIC ADJUSTMENT IN A CONTINUOUS ENVIRONMENT,

Corporate Author : ILLINOIS UNIV AT URBANA ELECTRICAL ENGINEERING RESEARCH LAB

Personal Author(s) : Andrew, A. M.

Report Date : SEP 1965

Pagination or Media Count : 23

Abstract : Control devices, biological or otherwise, which are able to adjust their own internal parameters are discussed. It is shown that under certain circumstances the adjustment process must depend on experimental fluctuations superimposed on either the parameters or the control signals. The way is studied in which the effective fluctuations attributed to the parameters can best be computed from fluctuations of the control signals. A mathematical comparison is given of two ways in which a self-improving controller may operate, namely with and without an explicit model of the environment. For a simple control task the two are shown to be almost exactly equivalent. (Author)

Descriptors :   *LEARNING MACHINES, ADAPTIVE SYSTEMS, CONTROL SYSTEMS, AUTOMATIC, OPTIMIZATION, ARTIFICIAL INTELLIGENCE, MATHEMATICAL MODELS.

Subject Categories : Cybernetics

Distribution Statement : APPROVED FOR PUBLIC RELEASE