An alternative iterative scheme for robust mixed sensitivity optimization

In this paper an iterative algorithm for robust mixed sensitivity optimization is derived. Firstly, a model based controller R(z) is designed to achieve robust mixed sensitivity for the nominal plant model G(z, theta ). In the second step, a new identification experiment is performed using the prediction error identification method. The identification prefilter D(z) is computed such that the weighted H₂ norm of the error between the designed and achieved sensitivity functions S(z, theta ) and S(z) is minimized. Next, the entire procedure is repeated until a satisfactory performance level for the true plant is achieved. The effectiveness of the proposed iterative algorithm is evaluated in the sensitivity optimization problem of a flexible robot arm.