Particle Swarm Optimization Algorithm And Its Application

Particle swarm optimization (PSO) is a self-adapting evolutionary computation technique, which is inspired by social behavior of bird flocking and fish schooling. Recently, PSO algorithm has been gradually attracted more attention and is becoming very popular because it is simple in concept, few in parameters, quick in convergence and easy in implementation. It has successfully been applied in the area of function optimization, combinatorial optimization, image processing, signal processing, scheduling problem and neural network training, etc. However, both the theory and application of PSO are still far from mature. There are still massive questions to be worth studying.The dissertation focuses on the principles, theory and application of PSO, especially, an in-deep study on how to improve the performance of the conventional PSO algorithm, solving the problems such as traveling salesman problem (TSP), multiple sequence alignment (MSA) and image enhancement. The main achievements of this dissertation are summarized as follows.(1) An improved self-organized particle swarm optimization (SOPSO) is proposed, which is based on the concept of "swap operator" and "swap sequence". To alleviate the premature convergence of basic PSO, the improved algorithm applies the self-organizing inertia weight and acceleration coefficients in the interest of the diversity of population. Moreover, the mutation operator is introduced. In view of the concept of "swap operator" and "swap sequence", the improved SOPSO algorithm which can search in the discrete domain directly is designed to solve the TSP. The simulation results show that the improved SOPSO algorithm is effective.(2) In view of the success of genetic algorithm (GA) and simulated annealing algorithm (SA) on the TSP, the two improved hybrid PSO (HPSO) are designed to solve TSP based on the study of the mechanism of HPSO.①A linear-descending hybrid PSO (LD-HPSO) is presented which uses a new method to accept the worse solution. In this algorithm, the particles acquire the updating information through intercrossing with individual optimal pbest and global optimal gbest.②A new improved algorithm called Geese-inspired hybrid PSO (Geese-HPSO) is proposed using the flight of geese for reference. In this algorithm, the particles are sorted according to the historical classic value. Then each particle flies with the particle ahead itself. The individual optimal pbest is redesigned as the sorted particles, and the global optimal gbest is redesigned as the particle ahead its corresponding one of the sorted population. The particles acquire the updating information through intercrossing with the redesigned pbest and gbest. Those quicken the convergence speed and strengthen the convergence precision greatly. The simulation results show that the two improved algorithms are effective. Especially Geese-HPSO algorithm has higher convergence precision and can search in the global scope more effectively than some other algorithms.(3) A multiple sequence alignment (MSA) algorithm is designed based on the chaotic PSO (CPSO). The algorithm applies the chaos theory to initialize the population so as to overcome the premature convergence. That enhances the diversity and the ergodic property of the particles when they are searching. In view of the Sums-of-Pairs with affine gap penalties (SP) optimization model of multiple sequence alignment, the CPSO is applied in MSA. The alignment results show the algorithm is effective, it improves the alignment precision and ability.(4) Based on the analysis of inertia weightωand the maximal flying speed Vmax, the improved particle swarm optimization with contracted range of search velocity (CV-PSO) is proposed for the adaptive image enhancement. It combines with incompleteβoperator which containes all different kinds of typical transformation functions. The new algorithm is used for the basic and traffic images enhancement. Then compare its performance with that of basic PSO and other improved PSO. The results show that CV-PSO is effective and superior. Moreover, it is better than traditional histogram equalization method in visual quality.In a word, this dissertation makes an in-deep and systemic study on PSO algorithm focusing on its shortage and practical application. Some effective improved measures are proposed, which provide some reference for the future study.