Applying Multiobjective Evolutionary Algorithms to Dynamic Software Product Lines for Reconfiguring Mobile Applications (bibtex)
by Gustavo G. Pascual, Roberto E. Lopez-Herrejon, Mónica Pinto, Lidia Fuentes, Alexander Egyed
Abstract:
Mobile applications require dynamic reconfiguration services (DRS) to self-adapt their behavior to the context changes (e.g., scarcity of resources). Dynamic Software Product Lines (DSPL) are a well-accepted approach to manage runtime variability, by means of late binding the variation points at runtime. During the system execution, the DRS deploys different configurations to satisfy the changing requirements according to a multiobjective criterion (e.g., insufficient battery level, requested quality of service). Search-based software engineering and, in particular, multiobjective evolutionary algorithms (MOEAs), can generate valid configurations of a DSPL at runtime. Several approaches use MOEAs to generate optimum configurations of a Software Product Line, but none of them consider DSPLs for mobile devices. In this paper, we explore the use of MOEAs to generate at runtime optimum configurations of the DSPL according to different criteria. The optimization problem is formalized in terms of a Feature Model (FM), a variability model. We evaluate six existing MOEAs by applying them to 12 different FMs, optimizing three different objectives (usability, battery consumption and memory footprint). The results are discussed according to the particular requirements of a DRS for mobile applications, showing that PAES and NSGA-II are the most suitable algorithms for mobile environments.
Reference:
Applying Multiobjective Evolutionary Algorithms to Dynamic Software Product Lines for Reconfiguring Mobile Applications (Gustavo G. Pascual, Roberto E. Lopez-Herrejon, Mónica Pinto, Lidia Fuentes, Alexander Egyed), In Journal of Systems and Software, volume 103, 2015.
Bibtex Entry:
@Article{DBLP:journals/jss/PascualLPFE15,
  author    = {Gustavo G. Pascual and Roberto E. Lopez-Herrejon and Mónica Pinto and Lidia Fuentes and Alexander Egyed},
  title     = {Applying Multiobjective Evolutionary Algorithms to Dynamic Software Product Lines for Reconfiguring Mobile Applications},
  journal   = {Journal of Systems and Software},
  year      = {2015},
  volume    = {103},
  pages     = {392--411},
  abstract  = {Mobile applications require dynamic reconfiguration services (DRS)
	to self-adapt their behavior to the context changes (e.g., scarcity
	of resources). Dynamic Software Product Lines (DSPL) are a well-accepted
	approach to manage runtime variability, by means of late binding
	the variation points at runtime. During the system execution,
	the DRS deploys different configurations to satisfy the changing
	requirements according to a multiobjective criterion (e.g., insufficient
	battery level, requested quality of service). Search-based software
	engineering and, in particular, multiobjective evolutionary algorithms
	(MOEAs), can generate valid configurations of a DSPL at runtime.
	Several approaches use MOEAs to generate optimum configurations of
	a Software Product Line, but none of them consider DSPLs for mobile
	devices. In this paper, we explore the use of MOEAs to generate at
	runtime optimum configurations of the DSPL according to different
	criteria. The optimization problem is formalized in terms of a Feature
	Model (FM), a variability model. We evaluate six existing MOEAs by
	applying them to 12 different FMs, optimizing three different objectives
	(usability, battery consumption and memory footprint). The results
	are discussed according to the particular requirements of a DRS for
	mobile applications, showing that PAES and NSGA-II are the most suitable
	algorithms for mobile environments.},
  bibsource = {dblp computer science bibliography, http://dblp.org},
  biburl    = {http://dblp.uni-trier.de/rec/bib/journals/jss/PascualLPFE15},
  doi       = {10.1016/j.jss.2014.12.041},
  file      = {:Journals\\JSS 2015 - Applying Multiobjective Evolutionary Algorithms to Dynamic Software Product Lines\\Applying multiobjective evolutionary algorithms to dynamic software product lines for reconfiguring mobile appli-preprint.pdf:PDF},
  keywords  = {FWF P25289, FWF M1421},
  owner     = {AK117794},
  timestamp = {Mon, 27 Apr 2015 09:11:46 +0200},
  url       = {http://dx.doi.org/10.1016/j.jss.2014.12.041},
}
Powered by bibtexbrowser