Gain-sensitivity augmentation for near-optimal control of linear parameter-dependent plants

This paper develops a parameter-adaptive version of the steady-state linear quadratic Gaussian controller for plants with structured parameter dependencies. Both scalar and vector parameter perturbations are treated. The design method, referred to as gain-sensitivity augmentation, is based on approximating the optimal parameterdependent regulator and filter gain matrices with truncated Taylor series expansions. The coefficient matrices of the Taylor series expansions are referred to as gain-sensitivity matrices and are precomputed off-line. Parameter information, assumed to be available on-line by direct measurement or estimation, is used to adjust the gain matrices to near-optimal values.