An input-mapping based online learning sliding mode control strategy for discrete-time uncertain systems

In this paper, a novel sliding mode control law is designed via an input-mapping based online learning strategy for systems subject to uncertainty. In view of the design of classical sliding mode control usually assume that the uncertainty varies in a given range, which is conservative for systems with time-invariant characteristics. Therefore, an input-mapping based online learning strategy is integrated into the proposed method to reduce the conservativeness of the classical way. The novelty of this work lies in the utilization of historical online information to predict the future system trajectory and to compensate the sliding mode control law, and the combination coefficient is studied by solving an optimization problem. Then, the convergence rate of considered systems is accelerated. Ultimately, a ball and beam system is given to illustrate the effectiveness of the proposed algorithms.