Font Size: a A A

Research On Typical Swarm Intelligence Algorithms And Their Update Mechanisms

Posted on:2019-07-13Degree:DoctorType:Dissertation
Country:ChinaCandidate:L Y ZhouFull Text:PDF
GTID:1368330572458277Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
The swarm intelligence algorithm is a kind of global optimization algorithms with simplicity,flexibility and generality.They are efficient methods to complex optimization problems and have been widely used in many areas of science,engineering,economics,military and management.Since the basic interaction rules of swarm intelligence algorithms are usually simple,the learning ability of individuals is not strong enough.And the algorithms always produce a large amount of intermediate data during the iterative process,while the useful information contained in the data are not fully utilized.The problems affect the performance of swarm intelligence algorithms.In order to solve the complex optimization problems more efficiently,this thesis improves the performance of the algorithms by improving the individual interaction mechanism and by extracting the useful information from the intermediate data using the opposition-based learning and other methods.The main research work is summarized as follows.(1)First,a mathematical model of the sampling inspection path planning problem is-given.Then,in order to solve the problem effectively,a type of multi-heuristic information ant colony optimization algorithm(MACO)is proposed based on the property of the problem.The reciprocal of the distance between the source node and the feasible node is joined in the probabilistic formula for choosing the next feasible node as a new heuristic information.The convergence property of the MACO and the effect of the new heuristic information are analysized.The simulation experiments are done on nine cases.The results demonstrate the availability of the new heruistic information for the problem and good performance of the MACO in terms of the solution accuracy.Finally,the new parameter is also analyzed in detail and the parameter value for better quality and better stability is given.The proposed approach provides an efficient method to the sampling inspection path planning problem.(2)An accurate partially attracted firefly algorithm(PaFA)is proposed by adopting a partial attraction model and a fast attractiveness calculation strategy.The partial attraction model can preserve swarm diversity and make full use of individual information.The fast attractiveness calculation strategy ensures the information sharing among the individuals and improves the convergence accuracy.PaFA can maintain the simplicity of the firefly algorithm(FA).The proposed algorithm are tested on the CEC' 2013 suite and a real world optimization problem.The experimental results demonstrate the good performance of the proposed algorithm in terms of the solution accuracy compared with some state-of-the-art FA variants and bio-inspired algorithms.(3)A neighborhood centroid opposition-based learning strategy and a neighborhood centroid opposition-based particle swarm optimization algorithm are presented.First,the neighborhood centroid is used as the reference point for the generation of the opposite particle,absorbing the population search experience and maintaining diversity.Second,the contraction factor is used to expand the reverse search space,increasing the probability of finding a better solution.Experiments are conducted on eight typical benchmark functions,CEC' 2013 test functions and also on a practical engineering optimization problem.The results verify the effectiveness of the neighborhood centroid opposition-based learning and the competitiveness of the proposed algorithm.The influence of the topology of the particles is also analyzed.The conclusion is that different topologies have little influence on the performance of the propposed algorithm.(4)Based on the analysis of related work on the existing opposition-based learning,a centroid opposition-based learning with a two-point full crossover is proposed by adopting a centroid opposition computing for considering the search information of population and a two-point full crossover for making the best use of the information in the original solution and its opposite.Then,the learning strategy is incorporated into the partially attracted firefly algorithm.The experiments are conducted on the CEC' 2013 benchmark suite and a real world problem.The proposed algorithm is compared with some state of the art FA algorithms and other up-to-date opposition-based evolutionary algorithms.The experimental results demonstrate the effectiveness of the learning strategy and the better performance of the proposed algorithm.(5)A novel learning strategy named Orthogonal Opposition-Based Learning(OOBL)is proposed and integrated into FA.In OOBL,first,the opposite is calculated by the centroid opposition,making full use of the population search experience and avoiding depending on the system of coordinates.Second,the orthogonal opposite candidate solutions are constructed by orthogonal experiment design,combining the useful information from the individual and its opposite.The proposed algorithm is tested on the standard benchmark suite and compared with some recently introduced FA variants.The experimental results verify the effectiveness of OOBL and show the outstanding convergence accuracy of the proposed algorithm on most of the test functions.
Keywords/Search Tags:Swarm intelligence algorithm, Update mechanism, Opposition-based learning, Orthogonal experiment design
PDF Full Text Request
Related items