首页> 外文会议>International Conference on QiR;International conference on civic space >Multi objective optimization based intelligent agent for NPC behavior decision
【24h】

Multi objective optimization based intelligent agent for NPC behavior decision

机译:基于多目标优化的NPC行为决策智能代理

获取原文

摘要

The main actor of the game is based on non-playable character (NPC) behavior to respond the environment based on artificial intelligent method. This research simulates the behavior of buyer-seller agent on purchasing computer goods in computer game. The buyer agent has price and specification variable which is reacted in satisfaction factor of agent. The seller agent has price and profit variable which is took effect in Join Utility (JU) of agent. In this case, there is usually no single optimal solution, but a set of alternatives with different trade-offs. This research describes buyer-seller agent behavior by multi objective optimizations approach using Multi Objective Evolutionary Optimization (MOEA) Non Sorted Dominated Genetic Algorithm II (NSGA II). NSGA II provides pareto fronts value to the minimum and maximum functions. Based on simulation result., we generate 3 kinds of scenarios of buyer and seller agent. First, the seller agent with profit oriented behavior provides the value of JU twice from the buyers function. Second, the seller agent with customer oriented behavior provides balance JU from the buyer function. Third, the buyer agent with satisfaction oriented behavior. Stability results of simulation is evenly attained after the fifth generation with simulation parameters: chromosome/pop=1000, crossover probability (pc)=0.9, mutation probability (pm)=0.005, index of distribution crossover (ηc)=20., index of distribution mutation (ηm) =20, value of pool=pop/2 and number of tour=2.
机译:游戏的主要演员基于不可玩的字符(NPC)行为来基于人工智能方法响应环境。本研究模拟了买方卖方代理在计算机游戏中购买计算机货物的行为。买方代理具有价格和规格变量,其在满足的代理人的满足因素中反应。卖方代理商有价格和利润变量,该盈利变量在加入效用(ju)的代理人中生效。在这种情况下,通常没有单一最佳解决方案,而是一组具有不同权衡的替代方案。本研究描述了使用多目标进化优化(MoEA)非分类主导遗传算法II(NSGA II)的多目标优化方法的购买商卖方代理行为。 NSGA II将Pareto Fronts值提供给最小和最大功能。基于仿真结果。,我们生成3种买方和卖方代理的方案。首先,具有利润导向行为的卖方代理商从买家职能提供两次ju的价值。其次,卖方以客户为导向行为的代理提供了来自买方功能的余额。第三,买方代理具有满意的面向行为。在第五代具有仿真参数之后均达到模拟的稳定性结果:染色体/流行音乐= 1000,交叉概率(PC)= 0.9,突变概率(PM)= 0.005,分布索引( η c)= 20。,分布突变索引(η m)= 20,pool&#x003d的值; pop / 2和巡回次数和#x003d; 2。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号