A self-tuning minimum variance controller based on multivariable multisample models

A minimum-variance self-tuning control algorithm applicable to multivariable systems represented by multisample input-output models is proposed. The theoretical development of the algorithm is analyzed in detail, and its performance is compared with that of an linear quadratic Gaussian (LQG) algorithm through practical control experiments using a micromachine system. As the minimum variance algorithm is implicit, it demands less computational load to evaluate the control action than the LQG algorithm. Consequently, the minimum variance algorithm seems to be more suitable for use with faster systems