Model free optimal control of unknown nonaffine nonlinear systems with input quantization and DoS attack

This paper is devoted to model free optimal control of unknown nonaffine nonlinear systems with input quantization and DoS attacks. Without model information, the system is presented by a modified compact form utilizing the quantized and attacked input. With this presentation, an optimisation criterion is used to approximate the unknown pseudo partial derivative parameter. Resorting to the adaptive dynamic programming approach, a single neural network-based weighting estimation law with variable learning rate is constructed to approximate the optimal cost function. Based on the approximated parameter and cost, an optimal control law is derived by applying the stationary condition. Sufficient conditions are established to make weighting approximation error and system state be uniformly ultimately bounded. The validity of the proposed model free optimal strategy is verified by illustrative examples.