摘要
影响图是贝叶斯网络的扩展,利用影响图可以有效解决智能Agent利用经验学习和行为选择2个基本问题。RoboCup是一个典型的多Agent系统,其中的球和球员都可以看成是Agent。文章给出一种基于影响图的决策方法,并将其应用到RoboCup中,从而实现了球员截球过程中行为的动态调整。
Influence Diagrams(IDs) are the expansion of Bayesian Networks(BNs),and it is effective to solve the two basic problems of intelligent agent by means of IDs—one is learning based on agent's observed history behavior,and the other is action selection. RoboCup is a typical multi-agent system, in which the ball and the players are regarded as agents. In this paper,a new decision-making method which is based on IDs is presented and applied to RoboCup to achieve the real-time adjustment of the action of a player in the course of intercept.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
2004年第4期388-391,共4页
Journal of Hefei University of Technology：Natural Science
基金
安徽省自然科学基金资助项目(03042305)
