IDENTIFICATION OF MODEL PARAMETERS AND ASSOCIATED UNCERTAINTIES FOR ROBUST-CONTROL DESIGN

The integration of system identification and robust control is considered. The identification algorithm is an extended Kalman filter, and the robust control algorithm is based on Petersen-Hollot's bounds (modified for random correlated parameters). The identification and control problems are coupled because the inputs for the identification experiment are selected to optimize the robust control performance. The optimization problem, interpreted as a form of Riccati equation control, is solved by exploiting the analytical properties of the Riccati equation in a nontraditional manner. The result appears an equivalent quadratic-linear boundary-value-problem, which allows a convergent numerical solution. An effective numerical algorithm is also offered for solving the extended Kalman filter equations in high-dimensional modal test problems. The algorithm is based on block decomposition of the modal state-space model. The developed approach is applied to the Middeck Active Control Experiment (MACE) testbed. MACE is an MIT STS flight experiment scheduled for launch in 1994.