Multidimensional Taylor network optimal control of SISO nonlinear systems for tracking by output feedback

Abstract

SummaryIt is difficult or even impossible for the existing nonlinear system control methods to realize the real‐time tracking control of nonlinear systems in the presence of modeling errors by output feedback. For that sake, multidimensional Taylor network (MTN) optimal control method is proposed for the real‐time optimal tracking control of the SISO nonlinear constant system with modeling errors as such. On the basis of the original MTN, the control input item is added to the nonlinear dynamic model, which is then termed as the MTN optimal controller (MTNOC). Its initial parameters are trained by the conjugate gradient method and the minimum principle, according to the calculated optimal input and output of the controlled object in open‐loop status. The propagation learning algorithm is adopted to improve the MTN product term weights and eliminate the modeling errors during the real system adjustment process. Simulation results show that the MTNOC promises a high response rate. Despite the errors in the modeling process, MTNOC manages to stabilize the system for tracking the desired output. The experiment of overlaying an additional signal on the input signal proves MTNOC to be of excellent tracking characteristics, capable of inhibiting the disturbance signal to some extent. In short, by realizing the optimal closed‐loop tracking control of the SISO nonlinear constant system by output feedback, MTNOC guarantees its real‐time control accuracy, dynamic performance, robustness, and anti‐disturbance capability.