One-step Optimal Correction for Input Saturation in Discrete Model-based Controllers

A one-step optimal method for compensating for any form of input saturation in discrete linear controllers is developed. The algorithm is straightforward to implement as an analytical correction operating on the difference between the past calculated control inputs and the inputs actually implemented. A further correction for multivariate controllers is necessary to simultaneously adjust the remaining control inputs if some inputs saturate The correction is valid for minimum variance and one-step optimal LQG controllers It can also be applied to Internal Model or Dahlin controllers. The algorithm is illustrated by simulation on several SISO and MIMO examples.