Application Research Of Improving Algorithm For PSO Based On Level Set

Different from the traditional optimization algorithm, intelligent optimization algorithm has greate advantage in the seeking global optimal solution of some engineering problems, so more and more people care about it recently. Although as important partment of Intelligent optimization algorithm: simulated annealing, ant algorithm, genetic algorithm and particle swarm optimization algorithm have different optimization progress, distinct characteristics and appear late, they have played a great role in combinatorial optimization and function optimization problem especially in NP-hard problem. At present many scholars engage in research of the four algorithms, including design parameters and theoretical research. Meanwhile, the trend of mergence of these optimization algorithms is more and more obvious.As one kind of intelligent optimziation algorithm, particle swarm optimization algorithm is preferred by so many scholars with its simple frame, few variables setting, parallel compute and rapid convergence. Recent years many literatures have been written about it.In generally,in the evolution of PSO swarm, all the swarms adjust these positions by two values: global best which has best position in the whole swarms in each generation and individual best which has best position in history for every swarm, so these two values are very important in this algorithm.This paper mainly disuss how to choose these two values with level set.Level set method has been usesally used in dealing with image problem including extract the edge of image and figure since it appeared .We also see its application on a new field: genetic algorithm ,thus the main ideal of this paper is its application on particle swarm algorithm. Firstly, we refer an important conception of level set, with the fit value of each particle swarm in every generation to construct level set. Then two kinds of improving method (global-best level-set based particle swarm optimization and individual-best level-set based particle swarm optimization) are referred, mutation operator and strategy of keeping optimization methods are also used to improve its robustness and convergence ability. The ability of new algorithms are tested in next chapter, several theory proofs are refered about new algorithms and some experiments are done, Compared to the result of experiment , we can make some improvement on the new algorithms .The contrast of these two methods are also added. The whole process of this two new algorithm can be seen as a fusion between GA and PSO.Lastly, the new algorithms are used to solve some practical issues such as disk load balance problem, changing the formula of new algorithm according to character of partical problem and the result is good.