Abstract: In order to overcome the shortages of Simulated Binary Crossover（SBX） operator, convergence speed and population diversity of NSGA-II, this paper applies the opposition-based learning mechanism?to the initialization and evolution process of NSGA-II algorithm. In addition, the paper introduces an improved arithmetic crossover operator as well. The convergence and diversity of the proposed algorithm on the series of ZDT test benchmarks are evaluated?and the results show that the improved NSGA-II algorithm is better than the traditional NSGA-II on converge speed, convergence and diversity. The paper applies the proposed algorithm to the optimization of satellite constellation design and the results indicate that the improved algorithm is very effective on this application.

Key words: multi-objective optimization, NSGA-II, opposition-based learning, satellite constellation

摘要： 针对NSGA-II算法中的模拟二进制交叉（SBX）算子以及NSGA-II在收敛速度及多样性保持方面性能的不足，将反向学习机制（OBL）应用到NSGA-II的初始化和进化过程中，并引入一种改进的算术交叉算子。ZDT系列测试函数在收敛性和多样性两个方面的评价结果表明，改进的NSGA-II算法在收敛速度、收敛性和多样性上优于NSGA-II算法。将改进的NSGA-II算法应用于卫星星座优化设计中，仿真结果表明改进的算法在卫星星座优化设计中比较有效。

关键词: 多目标优化, NSGA-II算法, 反向学习, 卫星星座