Genetic Algorithm Applications to Optimization Modeling

Genetic algorithms (GAs) are stochastic search techniques based on the concepts of natural population genetics for exploring a huge solution space in identifying optimal or near optimal solutions (Davis, 1991)(Holland, 1992)(Reeves & Rowe, 2003), and are more likely able to avoid the local optima problem than traditional gradient based hill-climbing optimization techniques when solving complex problems. In essence, GAs are a type of reinforcement learning technique (Grefenstette, 1993), which are able to improve solutions gradually on the basis of the previous solutions. GAs are characterized by their abilities to combine candidate solutions to exploit efficiently a promising area in the solution space while stochastically exploring new search regions with expected improved performance. Many successful applications of this technique are frequently reported across various kinds of industries and businesses, including function optimization (Ballester & Carter, 2004)(Richter & Paxton, 2005), financial risk and portfolio management (Shin & Han, 1999), market trading (Kean, 1995), machine vision and pattern recognition (Vafaie & De Jong, 1998), document retrieval (Gordon, 1988), network topological design (Pierre & Legault, 1998)(Arabas & Kozdrowski, 2001), job shop scheduling (Özdamar, 1999), and optimization for operating system’s dynamic memory configuration (Del Rosso, 2006), among others. In this research we introduce the concept and components of GAs, and then apply the GA technique to the modeling of the batch selection problem of flexible manufacturing systems (FMSs). The model developed in this paper serves as the basis for the experiment in Deng (2007).