Empirical Optimization of Non-linear Control Systems with Random Inputs. Part I†

†Communicated by Dr. A. T. Fuller.

ABSTRACT

The problem of optimizing a non-liner control system with a random input is solved approximately by fitting functions to experimental results. These results represent systems optimized experimentally using an analogue computer simulation and an automatic optimizer, or ‘bill-climber’. The controller is represented by a truncated power series. Dimensional analysis of the controller plays an essential role in defining the most efficient interpretation of the experimental data. The limitations imposed on the method by the optimization accuracy and some experimental approximations are discussed in connection with an example of a second-order relay system. Quite good agreement with theory is obtained, and it is concluded that this empirical approach should be useful in studying both simple and complex non-linear control systems; further examples will be given in Part II.

Notes

†Communicated by Dr. A. T. Fuller.

‡Previously at Cambridge University.