Data-Driven Tracking Control for a Class of Unknown Nonlinear Time-Varying Systems Using Improved PID Neural Network and Cohen-Coon Approach

In this paper, data-driven tracking control based on improved proportional-integral-derivative (PID) neural network and Cohen-Coon approach (IPIDNN-CC) is proposed for a class of completely unknown nonlinear time-varying systems (NTVS). This hybrid algorithm only uses the input/output (I/O) data of NTVS online, and initial control parameters of IPIDNN-CC can be properly determined only via system open loop response. Thus, IPIDNN-CC algorithm can ensure that initial control parameters can be converged based on gradient descent algorithm. Compared with traditional PIDNN, adaptive scaling factor is employed in the proposed improved PIDNN and the problem of PIDNN weights initialization is solved. Moreover, the stability of closed loop system based on IPIDNN-CC algorithm is proven based on Lyapunov stability theory. The simulations are carried out to demonstrate the correctness and superiority of the proposed algorithm.