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

The integration of system identification and robust control is conside red. The identification algorithm is an extended Kalman filter, and th e robust control algorithm is based on Petersen-Hollot's bounds (modif ied for random correlated parameters). The identification and control problems are coupled because the inputs for the identification experim ent are selected to optimize the robust control performance. The optim ization problem, interpreted as a form of Riccati equation control, is solved by exploiting the analytical properties of the Riccati equatio n in a nontraditional manner. The result appears an equivalent quadrat ic-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 p roblems. The algorithm is based on block decomposition of the modal st ate-space model. The developed approach is applied to the Middeck Acti ve Control Experiment (MACE) testbed. MACE is an MIT STS flight experi ment scheduled for launch in 1994.