On the initial conditions in learning control

The effect of the initial conditions on establishing learning control schemes for nonlinear systems is studied in this paper. The quantitative boundedness of the output errors between the actual and the given trajectories is obtained in the presence of initial state errors when a so-called D-type learning control is used directly. The robustness of learning control law with respect to the existence of initial errors, fluctuations of system dynamics and measurement noises is discussed. If high-gain state feedback to the open-loop systems is employed before the learning control loop is enclosed, then the output errors of the system will be independent of the size of the initial state errors at every operation.<
>