Research On Genetic Algorithms Based On Species Evolution

Genetic algorithm,as a bionic intelligence optimization technology, which can overcome the drawback of traditional gradient search technique, has been widely used in many fields of engineering science. It has more advantages over traditional optimization methods based on calculus theory when solving problems with global search, complex design and complicated target function optimization. It also has advantage of easy use. It has been a research hotspot of computational intelligence.The theories of simple genetic algorithm and improved genetic algorithms have been analyzed, and Darwin's Evolutionism and Mendel's genetics are taken as the guiding thought to imitate the methods and phenomena of ecology. In combination of some technical means in bioengineering and some theories and methods of the other subjects, simple genetic algorithm has been improved to increase the convergence precision and convergence rate. The concrete contents of this dissertation are as follows.The problem of species breeding is researched. Cross breeding genetic algorithm, mutation breeding genetic algorithm and selection breeding genetic algorithm are presented, which imitate different breeding technique in bionic engineer. The advantages and disadvantages of different breeding genetic algorithms are analyzed and compared by means of the results of the complicated function with minimum or maximum global optimal solution solved by the algorithms.The problem of species evolutionary environment is researched. Niche genetic algorithm illustrates that the species are reciprocity and interaction with one another rather than isolation. Penalty niche genetic algorithm is of excellent properties, which is often used to solve multi-modal function. But how to determine the niche distance hinders the algorithm application. Penalty niche genetic algorithm based on good character seeds is proposed, which uses simple genetic algorithm to preliminary exploration. Simple genetic algorithm runs several times to obtain good character seeds set. The location among the seeds can guide determining the niche distance.It is important to use prior knowledge when non-deterministic polynomial completeness problem in combinational optimization, such as travelling salesman problem, is to be solved. Subsection method based on good character seed is presented. Using the idea of strategy of'dividing-and-ruling'geographical space, the spatial search problems for optimum solutions through brute-force methods can be converted into problems of selection in limited projects. The incomputable problems in time can be solved in satisfying way by means of spatial disaggregation. The satisfactory solution can be obtained by rationally divided space according good character seed.The problem of species pleiotropy is researched. Genetic algorithm based on pleiotropy, which simulates pleiotropy of biology, is proposed. Nonlinear function between the genotype and phenotype or one-to-many mapping relationship is built, which breaks out one-to-one mapping in traditional genetic algorithm. The results show that pleiotropy has an advantage to maintain the diversity of the problem space so run rate can be improved by used small population size.The main innovate points of this dissertation are as follows.1. Cross breeding genetic algorithm, mutation breeding genetic algorithm and selection breeding genetic algorithm are presented, which imitate different breeding technique in bionic engineer.2. Seeking extreme point method is applied to obtain enough information to guide determining the niche distance, which is the key parameter of penalty niche genetic algorithm.3. The technique of open routing optimization and the technique of overlapping segments are used to solve the problem of end points of segments connection and optimization.4. Genetic algorithm based on pleiotroy is proposed, which builds nonlinear function between the genotype and phenotype or one-to-many mapping relationship, breaks out one-to-one mapping in traditional GA, and simulates the complex relationship between the biological genes and phenotype, effectively increasing the diversity of population.