求解约束优化问题M-精英协同进化算法

提出了一种适用于约束优化问题的协同进化算法.该算法旨在模拟人类社会中团队的组建及其协作方式,并强调精英人才对团队建设的推动作用.算法将整个种群分为精英种群和普通种群,围绕各个精英来组建团队,使精英种群带动普通种群,进而带动整个种群不断进化.组建团队过程中,不同精英之间采用协作操作,精英对普通种群成员进行引导操作,其中协作操作和引导操作由若干交叉或变异算子的组合所定义.使用静态罚函数法将约束优化转化为无约束优化,利用13个约束优化测试函数对算法进行了测试.仿真实验和参数分析结果表明,该算法寻优精度高,算法稳定,运行时间少,其性能优于组织进化算法,能够有效解决复杂的约束优化问题.

M-Elite coevolutionary algorithm for constrained optimization

An algorithm named the M-Elite Coevolutionary Algorithm (MECA) is proposed for constrained optimization problems. The algorithm simulates the establishment and cooperation of teams in the human society with emphasis on the important role of elites in teams. The whole population is divided into two subpopulations, i.e., elite population and common population, with the former leading the latter to evolve so as to accelerate the evolution of the whole population. During this process several teams are established, each elite individual acting as the core of each team. The cooperating operation is implemented between the different elite individuals, and the common individuals are led by the elite individuals in the leading operation during the process of team establishment. The above cooperating operation and the leading operation are defined by the different combinations of several crossover operators or mutation operators. Static penalty functions are used to transform a constrained optimization problem into an unconstrained one, and tests are made on 13 benchmark problems. Experimental results and parameter analysis show that the MECA can obtain a high solution quality with less runtime and is robust and that it obtains better performances than Organizational Evolutionary Algorithm, and can solve difficult constrained optimization problem.