Adaptive Probabilities of Crossover and Mutation in Genetic Algorithms Based on Cloud Model

Traditional genetic algorithms (GAs) easily get stuck at a local optimum, and often have slow convergent speed. A novel adaptive genetic algorithm (AGA) called cloud-model-based AGA (CAGA) is proposed in this paper. Unlike conventional genetic algorithms, CAGA presents the use of cloud model to adaptively tune the probabilities of crossover p_c and mutation p_m depending on the fitness values of solutions. Because normal cloud models have the properties of randomness and stable tendency, CAGA is expected to realize the twin goals of maintaining diversity in the population and sustaining the convergence capacity of the GA. We compared the performance of the CAGA with that of the standard GA (SGA) and AGA in optimizing several typical functions with varying degrees of complexity and solving travelling salesman problems. In all cases studied, CAGA is greatly superior to SGA and AGA in terms of robustness and efficiency. The CAGA converges to the global optimum in far fewer generations, and gets stuck at a local optimum fewer times than SGA and AGA