An investigation on noise-induced features in robust evolutionary multi-objective optimization

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalpeer-review

25 Scopus Citations
View graph of relations

Author(s)

Detail(s)

Original languageEnglish
Pages (from-to)5960-5980
Journal / PublicationExpert Systems with Applications
Volume37
Issue number8
Publication statusPublished - Aug 2010
Externally publishedYes

Abstract

Multi-objective (MO) optimization is a challenging research topic because it involves the simultaneous optimization of several complex and conflicting objectives that requires researchers to address many issues which are unique to MO problems. However multi-objectivity is only one aspect of real-world applications and there is a growing interest in the optimization of solutions that are insensitive to parametric variations as well. In order to evaluate the capability of MO evolutionary algorithms (MOEAs) to find robust solutions, it is important to employ suitable test functions. In this paper, empirical studies are conducted to examine the suitability of existing robust test functions. Results suggest that these test functions have a bias towards the region where the robust solutions lie, rendering it difficult to assess the true capability of MOEAs. Motivated by such a finding, we present a framework for the construction of robust continuous MO test functions characterized by different noise-induced features. These noise-induced features can pose different difficulties to the optimization algorithms. A fitness-inheritance scheme is also presented and incorporated into two well-known MOEAs. Empirical analysis of the proposed robust MO test functions reveals that some noise-induced features present greater challenges to robust MOEAs as compared to existing robust test functions. In addition, the vehicle routing problem with stochastic demand (VRPSD) is presented as a practical example of robust combinatorial MO optimization problems. The work presented in this paper should encourage further studies and the development of more effective algorithms for robust MO optimization. © 2010 Elsevier Ltd. All rights reserved.

Research Area(s)

  • Evolutionary algorithms, Multi-objective optimization, Robust solutions, Robust test functions