Improved PSO Algorithm Study And Its Application On Network Routing

Particle Swarm Optimization (PSO) is a random search method based on the biological community's preying. It is simple, robust and is fit to be used in the parallel computation. PSO is very effective in global search, and has been successfully applied in many aspects, but it also has some disadvantages such as easily to be premature, bad local search capability. This thesis analyzed the basic PSO and its modified algorithm systematacially. The main works are as follow:Firstly, in order to overcome the disadvantage that the basic PSO could be easily trapped in the local optimum. This thesis proposed a modified PSO. This algorithm divided the whole swarm into two groups. One group flies towards the globally best position found so far, the other (mutated particles) flies in the opposite direction. Thereby, the diversity and exploration of the space are increased, and the ability to break away from the local optimum is improved. Moreover, how the mutation rate would affect the result is also discussed.Secondly, this algorithm is applied to the PID controller tuning and the industrial design with restrictions. The results are fairly well.Thirdly, by far, the study on the discrete PSO is relatively not popular. When the basic PSO solving traveling salesman problem (TSP), it has many disadvantages, such as slow speed, low efficiency, hard to be expressed. In order to overcome these shortcomings, this thesis combined the PSO with GA (Genetic algorithm).Adopted the mutation and the crossover from GA, this combined algorithm can easily express the speed-location update formula. It divides the whole swarm into many groups, and makes them evolve separately for the primary election. The better local gene sequences obtained from the primary election are used for the global search. Simulation has been carried out on TSP and the result shows that the combined algorithm is robust and efficient.At last, the combined algorithm proposed in this thesis is applied to the IP network's QoS routing. The optimization target is to meet the QoS restriction, at the same time, the selected routing's cost must be minimum. Simulation result on a network shows that the algorithm is good.