For multi-objective optimization algorithms, the maintenance policy of external archive has a great impact on the performance of convergence and solution diversity. Considering the dilemma of large population and external archive, an improved strategy of external archive maintenance based on crowding distance is proposed, which requires less particle numbers and smaller archive size, resulting in the computation cost reduction. Furthermore, the information entropy of gbest is analyzed to emphasize the diversity improvement of non-dominant solutions and well-distribution on the Pareto-optimal front. Numerical experiments of benchmark functions demonstrate the effectiveness and efficiency of proposed multi-objective particle swarm optimization.