Some Aspects of Real-time Control of Linear Stationary Dynamic Systems

Abstract

The paper deals with the optimal positional control actions for linear discrete dynamic stationary objects. Methods of the real-time constructing of the implementations of optimal feedbacks are described based on reduction of the sequence of optimal programs to linear programming problems (LP).To solve the problem the duel method of LP with a long steps is used to correct supports, the parallelizing and acceleration of computations with the recurrent equation and methods of "set of weight" are applied. A real-time observation problem at uncertainty in initial states is considered to the linear dynamic object with discrete measurement devices which is reduced to the series of LP problems. Both of the problems are solved by the dual method of LP. To accelerate the computations of control actions it is suggested to use the parallelizing procedure.

Keywords

• linear stationary systems
• positional control
• dual methods of linear programming
• reccurent properties
• observation
• parallelizing procedure