Automatic Optimizing of a Poorly Defined Process-I

This paper is concerned with a consideration of the characteristics of various processes for which automatic optimizing might be employed. An appraisal is given as to the effect of these characteristics on the type of strategies or control logic to be employed. Further, an over-all description of a digital computer program is included which was developed and used to obtain quantitative results in this study. A number of strategies which were investigated in detail are then reviewed. The effect of such factors as constraints, measurement noise, and quantization are mentioned. Finally, the effects of noise are briefly examined, techniques to modify the strategies are discussed, and the results of one experiment described.