The Research And Simulation Implementation Of Differential Evolution Algorithm

Differential Evolution(DE) algorithm is a kind of evolutionary method for solving global optimization problems based on continuous variables, and also is a kind of bionic intelligent computation methods,which simulates Darwin’s Evolution theory “Selecting the superior and eliminating the inferior, the survival of the fittest", which have been proposed by Stom R and Price K since 1995. Three fundamental operations are included in DE algorithm:mutation, cross and selection. Differential evolution algorithm has characteristics of simple and easy to use, less controlled parameters, powerful search ability and so forth, compared with other evolution algorithms. Differential evolution algorithm is not only used for solving function optimization problems, but also have been broadly applied to many other fields, for instance, in the design of internet network routing protocol design, chemical engineering and formation reconfiguration.In the paper, we focus on the improvement performance of differential evolution.Sphere、Rosenbrock、Rastrigin、Griewank、Ackley and Noise are Benchmark functions, we take advantage of the six functions to compare three elementary differential evolution: DEI、DE2 and DE3(see 2.2.1 section in this article), which have different mutation strategies.Standard differential evolution algorithm DEI has the global search ability; DE2 could approximate the local optimal value; the property of DE3 is a compromise of DEI and DE3.Based on the complementary advantages of DEI and DE2 and different mutation strategies, we come up with the complex mutation strategy MDE2 with an alterable parameter^(/),(^(/)e(0,l)). The beginning value of is small, later being lager. In early stage,MDE2 search the optimal value by a decentralized way in a large scale. Later,MDE2 find out the optimal value using a centralized method on the key areas. In the given precision condition, numerical experiments show that the solving time and the numbers of iterations of modified differential evolution algorithm 2(MDE2) is less than MDEl which has a constant parameterS, such as S^O.5.In addition, we put forward multiple population differential evolution(MPDE) with two sub-populations. DEI and DE2 are employed in the two sub-populations respectively,furthermore,they contact with each other by migration operator for the purpose of coevolution. What’s more, we explored the immigration of MPDE. According to the experiments in this article, we find that the optimization will be better if we exchange the infonnation between the two sub-populations after a regular evolution interval. On the basis of the previous work, we apply DE、MDEl、MDE2、MPDE to train the weights and thresholds of BP neural network and solve the UAV’s path planning, obtaining an ideal simulation results.