Research And Application Of Particle Swarm Optimization

With out centralized controlling and global model, particle swarm optimization (PSO) provides a new method to find out the solution in complex distributed problem. There are two key issues in PSO algorithm. Firstly, achieving faster rate of convergence while searching at same accuracy. Raising the search precision while having comparable convergence speed. Secondly, based on this algorithm, effective method to solve practical problem should be studied. Moreover, research on application is helpful to improve performance of PSO.In the first place, for the problems of PSO proposed above, a number of studies are put forward in this paper. For PSO is of disadvantages of easy stagnation, poor convergence accuracy and low searching efficiency, some improvements on PSO are proposed. That is: improved PSO based on non-linear S function, chaos particle swarm optimization with isolation niches, improved particle swarm optimization with limited spread food information and study on adjusting velocity threshold of PSO based on multi-agent idea.In addition, for the second question, some researches are made. Particle swarm optimization is applied to aspects of intelligent traffic and mobile sensor network. For different practical problems, discrete and continuous models could be deduced through abstracting, simplifying and modeling to practicality. Therefore, The problems of traffic lights cycle controlling method in intelligent traffic and self-deployment of nodes in mobile sensor networks could be considering as optimization ones correspondingly. Using particle swarm optimization, effective way to solve problem could be realized.The main conclusions are as follows:(1) Based on S function, improved PSO adjusting parameters nonlinearly and adaptively is proposed in this paper. For the global searching ability, improvement could reduce the probability of stagnating in the local optimal value by 63%. It is of better performance than normal PSO.(2) Chaos particle swarm optimization with isolation niches could guarantee the diversity of individuals, which avoids prematurely convergence. Moreover, this improvement has ability to search around global optimal solution to enhance accuracy and speed. By comparison between normal PSO and the one embedding with isolation niches only, the latter has higher ability to solve complexity problems.(3) For optimal problems with multi-dimensions, improved particle swarm optimization with limited spread food information has better performance.(4) Explaining the particles behavior with multi-agents idea tentatively. Method adjusting velocity threshold of PSO based on multi-agents idea is put forward in this paper, which guarantees local and global searching ability.(5) PSO is applied to solve controlling problem of traffic lights cycle. By using existing transport facilities and not limiting vehicles, enhance roads capacity for traffic flows and reduce traffic congestion as far as possible. A true sense of the Green Wave Band could be formed. In that case, cities will have a safe, smooth traffic flow. Therefore, it could bring in effective traffic and society. (6) PSO algorithm is applied to self-deployment of sensor nodes. In consequence, the performance is enhanced.