Constraint aggregation method in a stochastic programming game problem (Q5928735)

The problem of searching for a value \(F_0\) and for the realization \(x^0\) of the minimum \[ \min_{x\in X}F(x), \quad F(x)=\int_Y \max_{z\in Z}f(x,y,z)\sigma dx, \tag{1} \] where \(X,Y,Z\) are compact sets in the Euclidean space, \(\sigma\) is a probability measure on \(Y\), the function \(f\) is supposed to be convex on \(x\) \(\forall y\in Y\), \(z\in Z,\) continuous and fullfilling the Lipschitz conditions on a set of variables, and integrable in relation to the measure \(\sigma\) \(\forall x\in X\), \(z\in Z\) is considered. The problem (1) is called a two-stage stochastic game problem with a recurrence. In operations research it corresponds to the problem of choice of the optimal strategy \((x^0)\) in presence of stochastic \((y)\) and indefinite \((z)\) factors. A reduction of problem (1) to a problem of maximization on a space of continuous functions and further to a semi-infinite optimization is proposed. The convergence of the method is investigated.