Study On Particle Swarm Optimization Algorithm With Multi-strategy

Swarm Intelligence (SI) is a relatively new evolutionary computation technique to simulate the collective intelligent behavior observable in the complicated system such as nature and society, which provides a new method and idea to solve complex, constrained, nonlinear and multi-minima global optimization problems. Particle Swarm Optimization (PSO) is an important branch of swarm intelligence. PSO is realized by the organic social behaviors, and by the cooperation and competition among the individuals themselves to search the optimum of the problem. PSO has been widely applied in image processing, pattern recognition, operational research and so on due to its simple concept, simplicity of implementation, less parameters to control and rapid convergence speed. However, Van de Bergh has proved that PSO cannot guarantee to converge on the global minimum. Aiming at the fatal limitation, keeping to the philosophy of PSO algorithm, the Quantum-behaved Particle Swarm Optimization (QPSO) has been proposed, which introduced the concept of quantum, built a Delta potential well model to simulate the learning inclination of particles and designed a method of controlling the parameters on global level. As the benchmark functions shown, QPSO has better performance than PSO.Firstly, the background and significance of the study are described in this paper. And then the Evolutionary Algorithm (EA) and Swarm Intelligence related to QPSO algorithm are introduced .And the theory of No Free Lunch, is simply explained.And then the differences and similarities among QPSO, PSO, EA and Genetic Algorithm (GA) are compared, which show the advantage of QPSO and the necessity of this research. And the further research on QPSO to develop is generalized.Secondly, PSO algorithm and the current research status of its improvements and applications are investigated. Then aiming to the potential dangerous characteristic of PSO algorithm, Particle Swarm Optimization with particles having quantum behavior based on Delta potential well (QDPSO) is proposed and its convergence and parameters control are analyzed in detail. To the vital parameter of QDPSO algorithm, creativity coefficient L, a method of parameter control on global level is further designed. Consequently the Quantum-behaved Particle Swarm Optimization (QPSO) is proposed.Thirdly, during the latter search period of QPSO algorithm, the particles are investigated to cluster together and it cann't exactly approach the optimal solution.So the strategy of division of work is proposed to increase its diversity of the swarm and improve the ability of global search.Then from the point of multimethod collaborative optimization algorithm,the algorithm based on QPSO is proposed, which effectively employs both the ability to jump out of the local minima in Evolutionary Programming and the capacity of searching the global optimum in QPSO algorithm.Next, QPSO algorithm is applied to the problem of integer programming. The experimental results show much advantage of QPSO to the traditional PSO. Moreover, QPSO and it's improved algorithm are applied to the Vehicle Routing Problem, which shows the better performance of QPSO.The main contributions of this paper are summarized and the further researches on QPSO are suggested at the end of this dissertation.