Solution set characterization of linear interval systems with a specific dependence structure (Q2641461)

Given interval matrices \({\mathbf A}^I\subset\mathbb{R}^{m\times n}\), \({\mathbf B}^I, {\mathbf C}^I\subset\mathbb{R}^{m\times h}\), and interval vectors \({\mathbf b}^I\), \({\mathbf c}^I\subset\mathbb{R}^m\), the author proves results of the following form: ``For certain \(A\in{\mathbf A}^I\), \(B\in {\mathbf B}^I\), \(C\in{\mathbf C}^I\), \(b \in{\mathbf b}^I\), \(c\in{\mathbf c}^I\) vectors \(x\), \(y\in\mathbb{R}^n\), \(z\in\mathbb{R}^h\) form a solution of the system \[ Ax+Bz=b,\quad Ay+Cz=c \] if and only if\dots'' where the dots are filled by some set of vector inequalities. The result represents a generalization of the well-known Oettli-Prager theorem for systems of linear interval equations prescribing now a particular sort of dependence structure. In the same way Gerlach's theorem is generalized by replacing the equality signs above by \(\leq\). Changing only the second equality above to an inequality finally results in a mixture of both problems. The proofs are essentially based on the duality theory in linear programming and on a relation between an optimization problem and a system of inequalities.