Coevolution Algorithm And The Application In Multi-agent System

Coevolutionary computation theory and application are current research hotspot. Evolutionary computation and coevolutionary algorithm have the advantages on bionics and are very suitable for solving many of the nonlinear problems which can not be handled by traditional methods for hard to establish accurate models. They have been widely used in the field of numerical optimization, machine learning, economic forecasts, robot path planning, navigation and intelligent control. In this paper, coevolutionary algorithm are focused as the target. The classic artificial intelligence problem of rounding up and TSP are used as the simulation environment. Some improved methods are proposed. Specific research works are as follows:First, In this paper,the aim of the research is to improve the efficience of genetic operator.The elite evolutionary algorithm which is based on distant hybridization was proposed on the distant hybridization theory and elitist strategy. At the initial stage, the algorithm population is divided into elite populations which are directly into the next generation without crossover and general populations who deliver the offspring by crossover based on the principle of distant hybridization. And the total offspring is made by the elite population and the offspring which is delivered by general populations. The simulation results show that the the opportunities for elite individual genetic,and the population diversity of genetic was enhanced by the algorithm.And the balance between in the depth and breadth of the search optimizing was acquired.The TSP experimental results show that the algorithm has reliable global convergence and faster convergence rate.Second,Advanced the Multi-fitness coevolutionary algorithm based on the feedback mechanism.This algorithm can improve the convergence rate. Finally, the model of robots rounded up simulation proved the effectiveness of the algorithm.