Optimal predictive control of linear discrete-time stochastic systems
| Autor: | Katsuhito Sakai, Takashi Yahagi |
|---|---|
| Rok vydání: | 1985 |
| Předmět: | |
| Zdroj: | Electronics and Communications in Japan (Part I: Communications). 68:27-36 |
| ISSN: | 1520-6424 8756-6621 |
| DOI: | 10.1002/ecja.4410681204 |
| Popis: | This paper considers the regulator problem for the linear discrete-time stochastic system represented by the input-output relation. The methods of optimal predictive control based on the prediction of output, when the output contains a delay in relation to the input, is described. In the control system, in general, there exists a delay inherent in the plant, and this delay apparently is increased with the decrease of the sampling period. There may also be produced a control delay due to the computation time when it is not negligible. From such a viewpoint, the predictive control is very important. In this paper, the input-output relation is represented by the equivalent state and output equations based on the observable canonical form, and a quadratic objective function containing output and control input is adopted. This choice of state variables is useful in that fewer computations are required in the derivation of the optimal control input, compared with other methods. By comparing the proposed method with a Kalman filter, either method can arrive at the optimal prediction, thereby minimizing the mean-square error under the specified condition. However, the proposed method is better in the number of computations when an adaptive control by digital computer is to be realized. For example, simulations are made by the proposed method from various viewpoints, indicating that satisfactory control can be realized by suitably selecting the objective function according to the property of the plant. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|