Rat Group Algorithm Based On Differential Evolution With Niching

Evolutionary algorithm has already become a hot topic for artificial intelligenceoptimization following the expert system, fuzzy mathematics and artificial neural network.And it is used in many scientific fields widely. Such as in automatic programming,machine learning, game strategy, cluster analysis and so on. It has solved some structureproblems successfully. For example: the aircraft structure design, construction of trussstructure and power grid structure. From the1990’s, the researches of the evolutionaryalgorithm have been carried out mainly in our country. It has been used successfully in theengineering technology of structure optimization and parameter optimization, for example:some people have used the mathematical statistics, operations research and the finiteelement numerical calculation method to optimize the parameters. Likes other sciencetechnology, the development of the evolutionary computation has experienced a gradualand growing development process.The research of optimization problem has a long history. In212-187BC, Archimedeswho is the Greek mathematician has proved the isoperimetric problem. Before the1950’s,the classical optimization problem of functional extreme value has been widely researchedby many scholars with the differential method and the variation method.Since the1950’s,the large-scale optimization problem begins to be solved by computer. When the problemis uncertain and multimodal nonlinear, it often can’t be solved well by the traditionaloptimization method. However, the evolutionary algorithm is used a black-box frameworktechnology. So that it has no requirements of mathematical expression. It is an effectivetool to solve complex optimization problems. For multimodal optimization problems, thetraditional heuristic search algorithms（such as simulated annealing method, artificialneural network）can’t avoid this problem that falling into local extreme point. Constructingan optimization algorithm which can find the global maximum points and the localmaximum value points as many as possible has become an open research problem.More and more scholars joined the research team of evolutionary algorithm andproposed a lot of effective optimization algorithm because of the amazing achievement ofthe evolutionary algorithms to optimization problem. The rat group optimization algorithmis a new type of simulated evolutionary algorithm, but it just used for the robot pathplanning in static environment. On the other hand, for multimodal function optimization problems and constrained optimization problems of continuous real functions, in theprocess of optimization it may move to a local optimal point by some evolutionaryalgorithms. That paper proposed a rat group algorithm by real number coding and nicheideology which achieves global optimization in the process of optimizing. And combiningwith the difference evolutionary algorithm, the experience factor is updated in time. In thisway it can improve the utilization of information, and enhance the exchange of informationbetween each individual. Therefore it avoids falling into local optimum effectively. Finallythrough the tests and simulations for some typical multimodal functions and continuousmultivariate unconstrained functions with more accurate evaluation index we prove theaccuracy and effectiveness of the proposed algorithm.