Adaptive primal-dual genetic algorithms in dynamic environments

  • Authors:

  • Hongfeng Wang;Shengxiang Yang;W. H. Ip;Dingwei Wang

  • Affiliations:

  • School of Information Science and Engineering, Northeastern University, Shenyang, China;Department of Computer Science, University of Leicester, Leicester, UK;Department of Industrial and Systems Engineering, Hong Kong Polytechnic University, Kowloon, Hong Kong;School of Information Science and Engineering, Northeastern University, Shenyang, China

  • Venue:
  • IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recently, there has been an increasing interest in applying genetic algorithms (GAs) in dynamic environments. Inspired by the complementary and dominance mechanisms in nature, a primal-dual GA (PDGA) has been proposed for dynamic optimization problems (DOPs). In this paper, an important operator in PDGA, i.e., the primal-dual mapping (PDM) scheme, is further investigated to improve the robustness and adaptability of PDGA in dynamic environments. In the improved scheme, two different probability-based PDM operators, where the mapping probability of each allele in the chromosome string is calculated through the statistical information of the distribution of alleles in the corresponding gene locus over the population, are effectively combined according to an adaptive Lamarckian learning mechanism. In addition, an adaptive dominant replacement scheme, which can probabilistically accept inferior chromosomes, is also introduced into the proposed algorithm to enhance the diversity level of the population. Experimental results on a series of dynamic problems generated from several stationary benchmark problems show that the proposed algorithm is a good optimizer for DOPs.