A convex approach for multicriteria decision making in hierarchical systems

In this paper hierarchical multicriteria optimization problems are addressed in a convex programming framework. It is assumed that the criteria are aggregated into a nonlinear function, which renders the problem nonseparable, in general. The projection of the problem onto the criteria space is used to obtain an equivalent separable problem, solved through a relaxation procedure implemented on basis of a multilevel structure. At the upper level of the structure, the decision making process involves the solution of a multicriteria problem formulated in the criteria space. The solution encountered at the upper level originates a lower level parametric optimization problem with separable structure that can be treated by standard coordination-decomposition techniques. The convergence of the overall procedure is ensured. The paper includes an application of the approach proposed for the control of dynamic systems with linear quadratic structure.