MOFNM - A method for optimum design of feedback systems

In this paper an innovative approach to the optimization of feedback control systems is presented. The proposed method uses ´banded´ prediction for computing the state trajectories, and converges to the optimum parameter vector by minimizing a family of I object functions in succession. This family is constructed, and ordered, in such a manner that the extremum of Ji is in the region of rapid convergence of Ji+1. Because of banded prediction, each of these functions retains a nearly quadratic behavior with respect to the vector of design parameters. These multiple object functions are minimized by the well known modified Newton method, hence the name, Multiple Object Function Newton Method (MOFNM). The success of the method is demonstrated by application to an example.