Multi-objective convex polynomial optimization and semidefinite programming relaxations (Q2038921)

A so-called hybrid method is proposed for determining efficient solutions to multiobjective optimization problems with convex polynomial inequality constraints. The first step of the approach relies on scalarizing the original multiobjective optimization problem, two kinds of representations of nonnegativity of convex polynomials over convex semialgebraic sets being discussed. Then two finite convergence results involving Lasserre-type hierarchies of semidefinite relaxations of the scalarized problem are provided, and finally is is shown that solving such hierarchies of semidefinite programming relaxations and checking a flat truncation condition yields efficient solutions to the original problem. Some examples illustrate the theoretical achievements.