Genetic State-Space Search for Constrained Optimization Problems

Abstract

This paper introduces GSSS (Genetic State-Space Search). The integration of two general search paradigms -- genetic search and state-space-search -provides a general framework which can be applied to a large variety of search problems. Here, we show how GSSS solves constrained optimization problems (COPs). Basically, it searches for "promising search states" from which good solutions can be easily found. Domain knowledge in the form of constraints is used to limit the space to be searched. Interestingly, our approach allows the handling of constraints within genetic search at a general domain independent level. First, we introduce a genetic representation of search states . Next, we provide empirical results which compare the relative merit of the introduction of constraints during the generation of the initial population, during the fitness calculation, and during the application of genetic operators. Finally, we describe some extensions to our method which came about when applying it to factory floor scheduling problems.