Condições de otimalidade para sistemas discretos no tempo com controle limitado pelo Estado

Abstract

Study of discrete optimal control problems in which the control is restricted by the state of the system. Initially, a rigorous self contained developnent of necessary and sufficient conditions is given through dynamic programming. This is followed by a developnent of the discrete maximum principle, presented in the manner of Cannon, Cullum and Polak. [6]. Examples are given to show that the maximurn principle usually does not hold in the case of state constrained controls and an "inclusion hypotheses" is given, demonstrating a class of problems where the usual maximum principle is valid. The Fritz John Theorem is then applied to obtain necessary conditions for problems with oonstraints of the form R(x,u) ≤ 0 and a modified maximurn principle is defined for this class. It is shown that linear positive systems of a type frequently encountered in economic systems satisfy the modified maximum principle. A number of detailed examples are given throug out the text.