CONTROLLER DESIGN AND QUALITY EVALUATION

NOVÁK, M.; BÖHM, J.

Abstract

Designing a controller for particular use requires a proper setting of its tuning knobs. The searched controller has to track prescribed setpoint properly while keep some constraints placed on its action variables. The tuning is well explored for three terms controllers. But, almost nothing can be found about tuning of LQ controllers. Tuning presented in this contribution is done for multidimensional adaptive LQG controller. Controller quality is evaluated by Monte Carlo approach using identified system model for performing off-line simulations. It is necessary to keep simulation as short as possible because lots of simulations has to be performed during tuning process. This contribution estabilishes on-line stopping rule for simulation run to achieve enough precise controller with the smallest possible computational demands.

Coresponding author e-mail: mira[at]utia[dot]cas[dot]cz

Session: Modelling, Simulation, and Identification of Processes