Adaptive dynamic programming based decentralized tracking control for unknown large-scale systems

This paper investigates decentralized tracking control (DTC) problem for unknown large-scale systems via adaptive dynamic programming. The DTC consists of local desired control and local tracking error control. By replacing the actual states of coupled subsystems in interconnection terms with their desired states, the norm-boundedness assumption can be removed. Based on local neural network (NN) observers, unknown subsystems can be identified. It helps to obtain the local desired control with corresponding desired states. By establishing a proper local cost function, the local tracking error control is derived via local critic NN and the identified input gain matrix. Consider the entire error caused by replacement, identification and critic NN approximation, an adaptive robust term is added to construct the robust local cost function that reflects the entire error, regulation and control simultaneously. Therefore, the asymptotic stability can be guaranteed for the closed-loop tracking system via Lyapunov stability theorem. To demonstrate the effectiveness of the proposed DTC, simulation on a hard spring connected parallel inverted pendulum system is given.