Gladwin, D. and Stewart, P. and Stewart, J. (2011) A controlled migration genetic algorithm operator for hardware-in-the-loop experimentation. Engineering Applications of Artificial Intelligence, 24 (4). pp. 586-594. ISSN 0952-1976

Preview

PDF controlled_migration.pdf - Whole Document Download (888Kb)

Abstract

In this paper, we describe the development of an extended migration operator, which combats the negative effects of noise on the effective search capabilities of genetic algorithms. The research is motivated by the need to minimize the num- ber of evaluations during hardware-in-the-loop experimentation, which can carry a significant cost penalty in terms of time or financial expense. The authors build on previous research, where convergence for search methods such as Simulated Annealing and Variable Neighbourhood search was accelerated by the implementation of an adaptive decision support operator. This methodology was found to be effective in searching noisy data surfaces. Providing that noise is not too significant, Genetic Al- gorithms can prove even more effective guiding experimentation. It will be shown that with the introduction of a Controlled Migration operator into the GA heuristic, data, which repre- sents a significant signal-to-noise ratio, can be searched with significant beneficial effects on the efficiency of hardware-in-the- loop experimentation, without a priori parameter tuning. The method is tested on an engine-in-the-loop experimental example, and shown to bring significant performance benefits.

Item Type:	Article
Additional Information:	In this paper, we describe the development of an extended migration operator, which combats the negative effects of noise on the effective search capabilities of genetic algorithms. The research is motivated by the need to minimize the num- ber of evaluations during hardware-in-the-loop experimentation, which can carry a significant cost penalty in terms of time or financial expense. The authors build on previous research, where convergence for search methods such as Simulated Annealing and Variable Neighbourhood search was accelerated by the implementation of an adaptive decision support operator. This methodology was found to be effective in searching noisy data surfaces. Providing that noise is not too significant, Genetic Al- gorithms can prove even more effective guiding experimentation. It will be shown that with the introduction of a Controlled Migration operator into the GA heuristic, data, which repre- sents a significant signal-to-noise ratio, can be searched with significant beneficial effects on the efficiency of hardware-in-the- loop experimentation, without a priori parameter tuning. The method is tested on an engine-in-the-loop experimental example, and shown to bring significant performance benefits.
Keywords:	Genetic algorithms, Hardware-in-the-loop, Migration, Response surfaces, Engines, ref15, refdoi
Subjects:	G Mathematical and Computer Sciences > G700 Artificial Intelligence
Divisions:	College of Sciences > Faculty of Science > Lincoln School of Engineering
Depositing User:	Paul Stewart
Date Deposited:	13 Jan 2011 18:37
Last Modified:	12 Apr 2013 10:02
URI:	http://eprints.lincoln.ac.uk/id/eprint/3832

Actions (login required)

View Item