Reduced-order models and controllers for continuous-time stochastic systems: an information theory approach

Reduced-order models and controllers for continuous-time stochastic systems are described. The reduced-order models are chosen to minimize the Kullback-Leibler information distance (KLID) between the outputs of the actual and reduced systems. An LQG controller based on a reduced-order system model is described. A second reduced-order controller is found to minimize the KLID between the closed-loop system outputs with the full and reduced-order controllers.