摘要
目前游戏中NPCs多目标行为进化是一个非常复杂的问题。对此建立了NPCs多目标优化的数学模型,并提出了一种NSGA-Ⅱ的改进算法——INSGA-Ⅱ。该算法在进行精英选择时,采用了基于K-均值聚类的方法联合了不同等级之间的个体进行集合划分,然后从不同的集合中选择下一代个体,从而更好地保持了种群的多样性。通过实例比较证明,在玩家和NPCs作战的游戏场景下,INSGA-Ⅱ能够得到NPCs复杂多目标控制问题的Pareto最优解,而且比NSGA-Ⅱ表现出更好的收敛性和多样性。
At present,NPCs multi-objective behavior evolution is a very complex problem in the games.In this paper,a multi-objective optimization model for NPCs is established and an Improved NSGA-Ⅱ algorithm-INSGA-Ⅱ is proposed.The algorithm is based on K-means clustering method which combines the individuals in different Pareto ranks for partition of sets,then selects next generation individuals from different clustering sets to maintain diversity of populations,finally,INSGA-Ⅱ and NSGA were compared in the specific game domain in which player and NPCs fight,INSGA-Ⅱ was capable of getting Pareto optimal solutions to NPCs complex multi-objective control problem and get better convergence and population diversity.
出处
《沈阳航空工业学院学报》
2010年第5期57-62,共6页
Journal of Shenyang Institute of Aeronautical Engineering
基金
辽宁省教育厅科学技术研究项目(项目编号:2008558)
中航一集团航空科学基金(项目编号:2008ZC54008)
沈阳市科学技术计划项目(项目编号:1091185-1-00)
