Abstract
This paper presents continuous-time Model Reference Learning Control (MRLC), which is a redesign of the Learnig Control (LC) from the viewpoint of the Model Reference Adaptive Control (MRAC). Through the same trials as LC, using a similar adaptive law as MRAC, MRLC decreases the control error between the outputs of the controlled object and the given reference model when the control can be applied repeatedly as the trial.
Unlike LC and MRAC, MRLC, is able to treat time varying linear systems even if the output of the reference model varies every trial and the time varing system is unknown without it's relative degree and stable invertibility. It is proved that the output error converges to zero through trials under these assumptions. A simple numerical simulation is presented for the verification of the analysis.
