Two Efficient Real-coded Genetic Algorithms for Real Parameter Optimization

This paper presents an efficient generation alternation model for real-coded genetic algorithm called rc-CGA. The most important characteristic of the proposed rcCGA model is the implicit self-adaptive feature in its crossover and mutation mechanism. By applying two crossover operators (BLX-α and UNDX crossover) in conjunction with Non-Uniform mutation to rc-CGA, respectively, we define two new real-coded genetic algorithms (rc-CGA+BLX+NUM and rc-CGA+UNDX+NUM). The proposed two realcoded genetic algorithms are compared with five existing real-coded genetic algorithms (MMG+BLX, MMG+UNDX, MMG+SPX, SGA+LX-NUM and JGG+REX) by simulating a set of 19 test problems available in the global optimization literature. The simulation results show that the rc-CGA is very efficient and that the rc-CGA+BLX+NUM performs quite well and outperforms other real-coded genetic algorithms for real-parameter optimization.