State Feedback Control Based on Twin Support Vector Regression Compensating for a Class of Nonlinear Systems

In this paper, we introduce a new twin support vector regression (TSVR) algorithm, which estimates an unknown function by approaching its up and lower boundary, the ending function is obtained by the mean of the two function. For the class of nonlinear systems composed by linear and nonlinear parts, we use TSVR with a wavelet kernel to estimate the unknown nonlinear part in the original system and to counteract it, and then a state feedback control is carried out to realize a close loop control for the compensated system. Simulation results show that the TSVR with the wavelet kernel has good approaching ability and generalization. The whole close loop system with a state feedback control is stable when the compensating errors satisfy certain conditions.