Neural dynamic sliding mode control of nonlinear systems with both matched and mismatched uncertainties

Mismatched uncertainty and chattering appear as two challenges in sliding mode control. To overcome the problem of mismatched uncertainty, multiple sliding surfaces with virtual inputs are proposed. Accordingly, we have proposed two new methods based on designed neural observer: sliding mode control (SMC) and dynamic sliding mode control (DSMC) methods. Although, the proposed SMC can significantly cope with the mismatched uncertainties, but it suffers from chattering phenomenon. The chattering problem can be removed in DSMC, because an integrator is placed before the system. This results in increased number of the system states. This new state can be identified with the proposed neural observer. Note that in both proposed approaches, the robust performance (invariance property) of system is reserved, even in the presence of mismatch uncertainties. Then, to have a valid comparison the proposed DSMC is also designed using loop transfer recovery observer (LTRO). This comparison shows the good performance of the DSMC based neural networks. Moreover, the upper bound of uncertainties is not used in SMC and DSMC controllers and also in the neural observer and LTRO, which is important in practical implementation. Finally, comparing the equations, one can see the simplicity of DSMC in concept and also in realization. (C) 2019 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.