Mean square error minimization in a nonstationary system

This paper gives a method for the analysis and synthesis of a controller to track a time-varying optimum operating point. In particular, the maximization of the average of a quadratic performance index of a general system is considered where not only the location of the maximum is time-varying, but also the shape of the performance index function. The optimization problem is reduced to that of a mean-weighted square error minimization one and Wiener filtering is employed. The controller, or filter, consists of linear integrators in conjunction with linear time invariant filters and instantaneous multipliers. Thus, an analogue computer can be easily adapted to control the system.