双赌轮选择遗传算法

提出了可以直接处理不等式约束条件的双赌轮选择遗传算法.为群体中的可行解和不可行解分别定义不同的适应度函数.可行解的适应度函数反映目标函数值的大小;不可行解的适应度函数反映约束条件的满足程度.建立2个赌轮分别选择可行解和不可行解.给出2个赌轮各自转动次数的公式,使可行解有更多的被选中机会.这样随着进化的进行,种群会自动向可行域内移动.这样就简单、直接地处理了不等式约束,并且允许初始种群全部为不可行解.另外,给出了一个改进的实数编码的两点交叉算子.比起典型的两点内插值型算术交叉算子,这个交叉算子具有更强的探测能力.算例测试表明双赌轮选择遗传算法的全局寻优能力较强,对不等式约束的处理简单有效.

Genetic algorithm with double roulette wheels

Double roulette wheels genetic algorithm was proposed to deal with the inequality constraints directly. Deferent fitness function was built for the feasible and infeasible individual in the population, respectively. The fitness function of the feasible individual reflected the objective function value; the fitness function of the infeasible individual reflected the degree of which the constraints were satisfied. Double roulette wheels was employed to select them respectively. A formula to decide the rotation times of each roulette wheel was given to make the feasible individual has greater probability to be selected than the infeasible ones. During the evolution, the individuals could move to the feasible region automatically. Then the inequality constraints were dealt with in a directive way and all infeasible initial population could be allowed. In addition, an improved real-coding copulation operator was also raised. This operator has more detecting ability than the classical two points inner interpolative operator. Examples show that the algorithm is not only an easy way to give global solution but also a simple and high effective method to deal with the inequality constraints.