The Circular Crowded Sorting (CCS) is adopted to improve NSGA-II algorithm for multi-objective optimization problem. CCS takes account of the affection to crowding distance of remained solutions, while one solution is eliminated. At first, the crowding distance of all solutions is computed in CCS, the crowding distance of extreme solutions is set to infinite; then the solution with least crowding distance is eliminated; the operator is cycled until the number of eliminated solution equal the appointed value. The time complexity of CCS is analyzed. Simulations indicate the improved NSGA-II algorithm has good performance of convergence and distribution.