摘要： 针对实时、多源、海量数据条件下用户所需信息的获取问题,提出一种面向对象的、基于多智能体协同的多源信息搜索模型,以对象为中心,在反馈循环搜索的过程中,完善对象描述模型并实现多源数据中关联对象信息的获取,提高多源信息获取的全面性和准确性。设计基于Q 学习的协同控制算法,针对马尔科夫对象与非马尔科夫对象给出相应的决策方法。实验结果表明,该协同控制算法比概率转移矩阵及概率统计算法具有更好的信息获取能力。

关键词: 多智能体, 信息搜索, 多源信息, 面向对象, Q 学习, 协同机制

Abstract: A new multi-source information search model based on multi-Agent collaboration is put forward to deal with the problem that under the real time,multi-source and huge information condition. Multi-Agent information search model centers around objects,builds the whole object model by cycling search,and gets the information that users care for. This model has higher intelligent and open-ended features, and it can make multi-source information searing more comprehensive and accurate. Q-learning-based collaborative control algorithm is proposed. The algorithm designs different decision-making methods for Markov objects and non-Markov objects. Experimental results show that the algorithm has better information search ability than probability transfer matrix and probability statistics algorithms.

Key words: multi-Agent, information search, multi-source information, object-oriented, Q learning, collaborative mechanism

中图分类号: 

• TP393