Modeling of parameter variations and asymptotic LQG synthesis

Conventional approaches in modern robustness and sensitivity theory are not adequate for the problems associated with parameter variation since the structure of parameter variations cannot be modeled properly or included in the synthesis procedure. A new modeling technique is proposed to handle a class of structured plant uncertainties in a direct way. The key is to treat deterministic parameter variations as an internal feedback loop so that the structure of parameter variations is embedded in its model. An asymptotic LQG design synthesis based on this modeling method is also presented. An important relationship between the structure of plant uncertainties and the LQG weighting matrices is obtained. This relationship clearly specifies the kind of parameter variations allowable for the LQG/LTR method.