Research On Some Problems Of Evolutionary Computation And Its Application

Evolutionary computation (EC), as a novel intelligent optimization technology, has been utilized extensively in engineering science. It has more advantages over traditional optimization methods when solving problems with global optimization, solving complex problems, and it is easier to use. Many branches of it appeared with development of EC, Genetic Algorithm (GA), Evolutionary Programming(EP), Particle Swarm Optimization (PSO), colony optimization, et. al are the common used algorithm. The research of EC always revolves two subjects: Firstly, to expanse the application domain of it; Next, to cause it more effective. The former is for the goal of designing and discovering effective of EC search strategies, solving the problem mat the past methods unable to solve or cannot solved effectively, and the latter is to revise and improve emphatically of the algorithm, causes it to be more effective.Some novel universal methods are designed to improve performance of EC with revolving the two subjects in the dissertation. In addition, some novel schemes are presented to solve some complex problems which we usual encounter with the proposed EC. The seven chapters of the dissertation are organized as follows:In chapter 1, some branches of EC and its history are introduced briefly firstly, and then actuality of improving the performance of EC is introduced, the three evolutionary methods which relate to the dissertation are mainly described. At last, the researching work and innovations points of the dissertation are briefly described.In chapter 2, two methods for improving performance of EC are designed. The first method is called average performance decreased temporarily evolutionary computation. in this method, partial individuals with low fitness in father generation are added in the new generation according to diversity. The relation between the number of chose individuals and diversity is given by a designed function, so, the algorithm has self-adaptive in some degree, and computation cost is small. The other method is called Ladder Evolutionary computation, compare to other variable population size methods, the algorithm has advantages in maintaining diversity and every individual can be evolved in adequate time. A self-adaptive method for adding and decreasing individuals is designed. The effectiveness of the idea is proved by using it in genetic algorithm and particle swarm optimization.In chapter 3, complex-valued encoding is used in Evolutionary computation.The individual is encoded by diploid with plurality in type of modules and angles, so the information contained in chromosome is expanded. In addition, a evolutionary programming with two steps chromosome is presented to optimize the structure and parameters of neural network and fuzzy rule bases is proposed in the chapter, and the effectiveness is expressed by modeling nonlinear system and fuzzy controller. The third part of this chapter is to design structure and parameters of radial basis function neural networks(RBFN) automatically with PSO. A novel flying method with self adaptive dimension is proposed to solve the problem that every particles should in the same dimension in conventional PSO.In chapter 4, motivated by high level regulation principle of endocrine system and neural system for individual's behavior, a PSO with endocrine regulation mechanism is proposed. The updating principle of hormone is the main object of the study, and an novel function which is fitting for hormone updating is presented, the updating equations for particles are modified. The effectiveness is demonstrated by optimization experiments of typical function and global path planning for robotic.In chapter 5, An PSO with "inverse population" cooperation is proposed, According to the different characters of tracking better position and exclude worse position, the two swarms have different behavior are used to improve the performance of PSO. The disadvantages of local convergent for swarm with absorbed behavior and not convergent for swarm with exclusion behavior because there is no leader of good information are avoided.In chapter 6, a novel method of designing centers of fuzzy rule bases offline and online based on combining maximum entropy principle and evolutionary computation are proposed. The inferential process of it is anaysised, some typical functions are used to test the effectiveness of the method, and the online method is used in complex movement tracking, the overflow is avoided in the process of determining maximum entropy principle.In chapter 7, a summary of the research conclusions and a discussion on the most promising paths of future research are also presented.