Assessment of an hybrid multi-objective pattern search filter method

Abstract

A hybrid multi-objective evolutionary algorithm (MOEA) for solving nonlinear multi-objective opti- mization problems that relies on a pattern search filter method is proposed. The aim is to reduce the computational time involved in solving expensive multi-objective problems by improving a subset of Pareto points. The proposed pattern search filter method relies on two components. Each entry in the filter aims to measure feasibility and optimality. The feasibility and optimality come directly from each single-objective nonlinear program problem that is associated to the multi-objective problem. Experi- ments carried out with a set of nonlinear multi-objective problems show that our pattern search filter approach is effective in reaching improved Pareto points. A comparison with other techniques known in the literature is presented.