Adaptive NN Control Without Feasibility Conditions for Nonlinear State Constrained Stochastic Systems with Unknown Time Delays.
Dapeng Li,Lei Liu,Yan-Jun Liu,Shaocheng Tong,C. L. Philip Chen
DOI: https://doi.org/10.1109/tcyb.2019.2903869
IF: 11.8
2019-01-01
IEEE Transactions on Cybernetics
Abstract:In the novel, an adaptive neural network (NN) controller is developed for a category of nonlinear stochastic systems with full state constraints and unknown time delays. The control quality and system stability suffer from the problems of state time delays and constraints which frequently arises in most real plants. The considered systems are transformed into new constrained free systems based on nonlinear mappings, such that full state constraints are never violated and the feasibility conditions on virtual controllers (the values of virtual controllers and its derivative are assumed to be known) are removed. To compensate for unknown time delayed uncertainties, the exponential type Lyapunov-Krasovskii functionals (LKFs) are employed. NNs are utilized to approximate unknown nonlinear functions appearing in the design procedure. In addition, by employing dynamic surface control (DSC) technique and less adjustable parameters, the online computation burden is lightened. The control method presented can achieve the semiglobal uniform ultimate boundedness of all the closed-loop system signals and the satisfactions of full state constraints by rigorous proof. Finally, by presenting simulation examples, the efficiency of the presented approach is revealed.