Abstract: In allusion to particle swarm optimization being prone to get into local minimum，an improved particle swarm optimization algorithm is proposed.The algorithm draws on the thinking of the greedy algorithm to initialize the particle swarm. Two swarms are used to optimize synchronously，and crossover and mutation operators in genetic algorithm are introduced into the new algorithm to realize the sharing of information among swarms.This paper tests the algorithm with a Traveling Salesman Problem with 14 nodes.The result shows that the algorithm can break away from local minimum earlier and it has high convergence speed and convergence ratio.