On a primal-dual interior point filter method

Abstract

Here we present a primal-dual interior point method that relies on a filter line search method to promote global convergence of nonlinear optimization problems. Each entry in the filter includes two components directly taken from the first-order optimality conditions of the problem. Primal feasibility, complementarity and dual feasibility measures are used to compose the components. The logarithmic barrier function is also used to promote convergence to minimizers. The algorithm is tested with a benchmark set of nonlinear problems.