Neural network based finite horizon stochastic optimal controller design for nonlinear networked control systems

Existing neuro-dynamic programming (NDP) techniques are not applicable for optimizing real-time NNCS with terminal constraints during the finite horizon. Therefore, a novel time-based NDP scheme is developed in this paper to solve finite horizon optimal control of NNCS. First, an online neural network (NN) identifier is introduced to approximate the control coefficient matrix. Then, the critic and action NNs are utilized to determine time-based finite horizon stochastic optimal control for NNCS in a forward-in-time manner. By incorporating novel NN weight update laws, Lyapunov theory is used to show that all closed-loop signals and NN weights are uniformly ultimately bounded (UUB) with ultimate bounds being a function of initial conditions and final time. Moreover, the approximated control input converges close to target value within finite time. Simulation results are included to show the effectiveness of the proposed scheme.