Ostrovs'ka Z. Adaptive control algorithms for multidimensional stochastic objects in presence of constraints on phase variables

In the dissertation control problem for the multidimensional stochastic object with unknown parameters and disturbances subject to constraints on variables and errors is considered. Structure of the optimal multi-step predictor with linear filter in the output is synthesized. For the predictor parameter tuning, the algorithm protected from covariance matrix wind-up is proposed. For state space model, the one-step ahead optimal adaptive control algorithm is obtained that guarantees implementation of various constraints on the state and control variables. Implementation of constraints is guaranteed by outer loop of the adaptive system that corrects both structure and parameters of the cost function. Adaptive control algorithm for multidimensional stochastic object subject to constraints on the input and output variables and errors is proposed. Using concepts of the multi-purpose optimization, adaptive control algorithm subject to constraint on the data processing time is obtained.