The author considers robust adaptive control of a continuous, linear, time-invariant plant in the presence of bounded disturbances and unmodeled dynamics. The novelty of the author’s control scheme lies in the use of two simultaneous estimators. Both estimators are used at every instant of time, in order to minimize plant identification error, and produce a filtered control for the plant. Plant parameter estimates are a convex combination of the individual estimator’s results. Then, using a switching rule that requires a minimum fixed duration, an estimation scheme is selected. The switching mechanism relies on minimization of a loss function of the time-integral of the square tracking error over a receding finite-time horizon.
The author derives his results for two different estimation strategies that have a regressor that is shared with the plant or separate regressors for the input and its derivatives.
A numerical example is provided to illustrate the author’s approach. A linear, time-invariant fourth-order continuous system is used; it could represent, for example, a cascade of four resistive-capacitive filters in an electrical circuit. The author’s approach shows improved transient performance attributable to switching between two estimation schemes.