Regression techniques in multivariate adaptive control systems

A control system is proposed which utilizes a multiple regression technique for optimal control of a multivariate nonstationary process. The concept underlying this system is the basing of each set of control actions on a current statistical description of the process, so that the adjustments performed will cause control of the output in an efficient manner. Adjustments are made to occur at discrete intervals. Descriptions of the applicable statistics and the computations required are presented, and the problem of choosing decision rules to relate control actions to process statistics is discussed. Several decision rules of practical importance are examined.