The adaptive dynamic programming based on the data - A total least squares approach

Focusing on the problem of the input and output data both contain measurement noise in linear time invariant system, this paper proposes that utilizing Tikhonov regularization of total least squares to solve the ill-poseness in process of adaptive dynamic programming. By applying the presented algorithm, the designed controller is obtained through the input and output data of the system, i.e. the method of controller designing has data-driven property. Traditional approaches only in allusion to the error in output data, if considering the ill-conditioned problem of the input and output data both contain noise disturbance, the algorithm of this paper can solve it and has robustness. Illustrative example indicates that the presented algorithm can give a good controller for a system, even when the measured outputs are polluted by noise. Compared with existing results, the presented algorithm has more effectiveness.