Research On Coevolutionary Based Hybrid Intelligent Optimization Algorithms

In real life,the optimization problems are involved in many fields.The emergence of various kinds of computational intelligence methods not only solve the optimization problem,but also provide a novel,unique and efficient way to solve the problem.However,with the complexity and diversification of optimization problems,a single computational intelligence method always presents some limitations in terms of quality,efficiency,convergence speed or global search ability.Therefore,combining different types of computational intelligence methods and constructing a hybrid intelligent optimization algorithm in an efficient,special way become more and more necessary.First and foremost,the principle,model,description and the implementation process of algorithm are introduced in detail by this thesis,as well as the convergence and complexity of the computational intelligence methods are analyzed in detail.Furthermore,the operation mechanism and implementation process of co-evolutionary algorithm are introduced in detail,and the computing method of fitness of competitive and cooperative co-evolutionary algorithm is given.Last but not least,this thesis analyses the research status of co-evolutionary algorithms and hybrid intelligent optimization algorithms.Combining the theory of computational intelligence with co-evolutionary algorithms and aiming at the problems based on other researchers,a series of research work have been done in this thesis.1.Based on the analysis of research status and the theory of co-evolutionary algorithm,this paper proposes a parallel co-evolutionary particle-ant algorithm which the parallel evolutionary model and the co-evolutionary idea used to combine PSO and ACO.The algorithm maintains two populations.Based on the idea of information migration and knowledge sharing,some rules are put forward to control the migration of individuals,so as to obtain the diversity of population and exchange the information among the populations,the co-evolution among populations realized finally.So that it is possible to solve the optimization problem of indecomposable single-target parameter Finally,the algorithm breaks the limitation which the single-target parameter can not be decomposed.2.Based on the cooperative strategy,this thesis puts forward an ingenious method.The method utilizes the strong positive feedback and robustness of ant colony algorithm to update the velocity and position of the particle swarm optimization,so as to increase the social attribute of particle.At the same time,the pheromone of ant colony algorithm is updated by the optimization result of particle swarm optimization,so as to increase the searching range of the algorithm.The performance of the algorithm is improved finally.3.Based on the idea of hybrid realization in stages,this thesis proposes a two-stage hybrid intelligent optimization algorithm which take the PCEPA algorithm as the first stage and the genetic algorithm as the second stage.Experiments prove that the algorithm has a faster convergence speed than the genetic algorithm,and a higher accuracy than the PECPA.Finally,the performance of the algorithm is verified by solving the TSP problem.4.When two-stage hybrid intelligent optimization algorithm is used to solve TSP,two new crossover operators are introduced in the genetic algorithm stage,this thesis proposes a new idea of hybrid crossover operator.The method greatly improves the performance of genetic algorithm,and stops the algorithm from falling into the local optimal solution early.