On robust optimization of two-stage systems (Q1424290)

A classical two-stage stochastic programming problem is a composition of two (outer and inner) optimization problems. The solution of the outer problem can depend only on an ``underlying'' probability measure while the solution of the inner problem can depend on the solution of the outer problem and on the realization of the random elements. Moreover, the solution of the outer problem is sought with respect to the mathematical expectation of the objective function that includes the optimal value of the inner problem. It is known that this type of problem can be employed for financial decisions. Of course, then a solution is sought with respect to a minimal risk (or to a maximal profit); it corresponds to the risk neutral approach. According to this fact a new type of problem has been suggested. In detail, the objective functions have been completed by a function corresponding to a variability measure, usually the variance of the second stage ``cost''. A new approach is called the robust optimization of two-stage systems. It is known that, the two-stage stochastic linear programming problem can be solved by a linear programming problem in the case of a probability measure with a finite support. Of course, a dimension of the new linear problem is large. The paper considers a robust optimization model with a discrete probability measure and a finite support. Evidently, this case corresponds to a scenario approach. The relationship between the original inner problem and the new approach is investigated in the paper. Furthermore, a possibility to employ (a variant of) the L-shaped decomposition algorithm for solving the new ``robust'' model is introduced. At the end of the paper, numerical demonstrations are presented.