An Output Matching Approach to Multivariable Linear Digital Control.

Abstract : A simple and direct state space approach to the digital control of multivariable linear system is discussed. Control is provided by minimizing the mean square error between controlled plant outputs and specified desired output trajectories at sampling instants. Systems of linear equations for digital control inputs result with solutions assuming a natural constant forward and feedback gain form. Optimal gains are determined using elementary results form linear systems theory and standard techniques from linear algebra. Numerical applications to examples in simple model following, digital redesign, and direct digital design are described. Partial state observability and the effects of sampling rate upon system performance are considered. Control smoothing through matching at multiple sampling instants is discussed. This strategy provides improved performance without additional computational expense or increased sampling rate. (Author)