Real Versus Binary Coding in Genetic Algorithms: A Comparative Study

Abstract

The paper presents a comparative study between binary-coded (Bga) and real-coded (Rga) genetic algorithms. In a traditional binary-coded genetic algorithm, binary sub-strings corresponding to each design variable are stacked head-to-tail to yield a binary string that represents a particular design. In contrast, the real-coded genetic algorithm does not make use of the binary representation, allowing for gene transformation operations to be conducted on the original real-valued representations of the design variables. The study describes the use of a real-coded genetic algorithm to solve optimization problems containing continuous design variables or a mix of continuous, integer and discrete design variables. An approach for extending real-coded genetic algorithms to problems involving integer and discrete design variables is proposed in this context. Two illustrative problems are solved using both the binary and real-coded genetic algorithm with the focus of exploring relative strengths and weaknesses of these competing methods.