Memory-Augmented Adaptive Control With Parameter Convergence Guarantee

A modified concurrent learning adaptive control, named memory-augmented adaptive control, is developed in this paper, which guarantees the convergence of parameter estimates without the requirement of persistent exciting condition and state derivative estimation. A stable low-pass filter is introduced into the error dynamics to filter out the state derivatives, circumventing the estimation of the unmeasurable state derivatives. And, the filtered basis matrix is also used in this adaptive control law to achieve the exponential convergence of the estimation error. Numerical simulations show that the proposed memory-augmented adaptive control performs better in tracking and estimation error convergence, compared to the standard adaptive control.