Identification with modeling uncertainty and reconfigurable control

The problem of obtaining reliable estimates of uncertainty in the parameters identified through a least-squares algorithm is discussed. Estimates based on a stochastic analysis, an analysis assuming bounded noise, and a sensitivity analysis are reviewed. The results are compared and illustrated using experimental data obtained on a DC motor. The need for methods of estimation of uncertainty is justified in the context of adaptive control, where robustness and transient performance are critical. In particular, the application to reconfigurable flight control is considered. Design tradeoffs for this application are discussed in detail and illustrated through simulations using two aircraft models