Udovenko S. Suboptimal control of multiconnected stochastic plants based on Bayesian approach

The thesis is devoted to solve scientific problem of creating new methods of suboptimal control for multiconnected stochastic plants based on Bayesian approach. The basic attention is given to questions of reduction for ARMAX and RARMAX-models in control systems and parameter tracking based on Bayesian estimation. The problems of recursive procedures design for parameter and state of stochastic linear and quadratic prediction models were considered. The complex of suboptimal control methods for quazistationary plants based on search of stochastic-equivalence, cautious and mixed strategies was proposed. The problem of suboptimal control for processes with various types constraints on the state and control variables was considered. An approach to decentralized control system design which are robust to interconnection disturbances between local subsystems, i.e. for systems retain guarantee asymptotical stability and suboptimalness was proposed and investigated. The dissertation results can be used in design of digital control systems for multiconnected quazistationary processes.