The data-based LQG control problem

Defines a data-based controller as one that can be synthesized using only knowledge of the plant input-output data, requiring neither a state space model nor a transfer function of the plant. In this paper the data-based LQG control problem is formulated and solved. Given finite Markov parameter data sequences (which can be computed from almost any input-output data) between the input-output and noise-output, the problem of data-based LQG control is to find the optimal control sequence which minimizes the quadratic cost function over some finite interval [0, N]. The authors show that a state space model is not necessary for this problem. Rather a finite sequence of input-output data is required to compute a finite set of Markov parameters. This data can be computed off-line, or on-line, on the previous control period [0, N] to prepare for a subsequent period [N, 2N]. A numerical example is given to illustrate the effectiveness of the data-based control theory