Research On Particle Swarm Optimization And The Application In FIR Digital Filter

With the quick development of computer science and technology, optimization technology and intelligence compute have obtained much concern. As one of swarm intelligence algorithms, particle swarm optimization(PSO) has got much attention and developed quickly. PSO is an random search algorithm which originates from bird’s foraging behavior, and has been an important optimization tool for solving science research and project optimization. So the research to PSO has profoundly scientific value and practical meaning either from the viewpoint of theoretical or applied research.To be specific, the thesis mainly discusses the issue from the following aspects:At the first part, after introducing the improved strategies of PSO learning parameters including inertia weight and learning factors, and analyzing the deficiency of previous method, the thesis presents a new PSO which utilizes inertia weight to control learning factors. The new PSO enhances the ability of cooperation and information sharing among particles by building a specific function relation between inertia weight and learning factor. Then three function relations are adopted including linear, nonlinear and trigonometric function for comparing algorithm performance. At last, the availability and superiority of improved PSO which is based on inertia weight and learning factors are proved by the simulation analysis to testing functions.Then, population diversity which is also a significant element of affecting algorithm performance will be discussed and analyzed. This part mainly researches the relation between population diversity and learning parameters by using a computing method of population diversity based on population centre that Shi and Eberhart presented, combined with multivariate function theory, random process theory and previous function relation between inertia weight and learning factors. Above research results will provide powerful guidance for preferably controlling population diversity and choosing algorithm parameters, provide theoretical gist for improved learning parameters of former, and lay a foundation for importing mixture inertia weightFinally, after introducing chaos optimization, simulated annealing and crossbreed PSO, a new mixture PSO(MPSO) which contains above three improved technologies is proposed. On the basis of the former, a mixture inertia weight combined with linear and exponential decrease is imported to MPSO. Then, after analyzing finite impulse response(FIR) digital filter’s design method, the thesis utilizes MPSO which is based on mixture inertia weight and learning fators of the thesis to design FIR low-pass and band-pass digital filter. MPSO’s feasibility and availability are demonstrated by optimization analysis and comparison with other corresponding mixture PSOs. In addition, the design results will provide reference value for mixture PSO’s application in other digital filters.This thesis mainly researches PSO’s parameters, the relation between population diversity and parameters, and its application on FIR digital filter. These researches not only can help researchers to be a better understanding of PSO, but also can contribute to select appropriate algorithm parameters as long as meeting practical problems and produce unexpected effect and meaning for national economy and the people’s livelihood.