Abstract
Switching control is used as a feasible control technique with low-resolution actuators, and it is modeled as a control system which restricts its input to discrete values. The control performance of this control besically depends on the length of the control period and switching surface. Long control period will cause large chattering, and short control period will cause redundant switching and energy consumption. The design of switching surface is difficult problem for nonliner plant.
In this study, firstly, a switching of discrete-valued control is considered as a timed discrete event. Secondly, propose a controller design method which realizes variable control period state feedback control by exploiting learning capability of neural network. The designed neural controller determines the control period and the switchng surface simultaneously. Finally, the usefulness of the proposed method is demonstrated through some numerical simulations.
