Abstract. We propose a new method for solving nonconvex semi-infinite problems by using a concave overestimation function of the semi-infinite constraints. At each iteration we solve a nonlinear programming problem locally which gives a feasible point, for certain problems the feasibility is so important than the optimality (e.g. in control systems design). If we decide to stop our algorithm after a finite number of iterations, we have an optimal solution or an approximate solution which is feasible.

Received: August 30, 2012

AMS Subject Classification: 65K05, 90C33, 90C34

Key Words and Phrases: semi-infinite optimization, global optimization, upper bound function, w-subdivision.

Download paper from here.