A new closed-loop identification technique for step-response model based on subspace approach

In this paper, a new identification technique for step-response model under closed-loop based on subspace approach is presented. It is well known that the lower block triangular Toeplitz matrix in subspace identification consists of impulse-response coefficients, these coefficients can be directly obtained from the collected data from experiments. The step-response model is constructed subsequently. It could be the key point for model-free predictive control. In order to estimate these coefficients in closed-loop circumstance, an orthogonal decomposition is introduced. For decreasing the variance and improving the performance of estimation, a parsimonious algorithm is used. Monte Carlo simulations are provided to illustrate the proposed method.