异构Agent协作的研究进展

Progress in Heterogeneous Agent Collaboration

首先在介绍多Agent协作的研究背景以及国内外研究现状的基础上,对多Agent理论及其协作机制分别进行了分类和总结;然后,在研究多Agent协作框架的基础上,分析了当前最具代表性的几种多Agent协作模型,如黑板模型、合同网模型、熟人模型、关系网模型以及承诺模型,阐述各模型特点并在动态性、异构性、灵活性和反应性等方面进行比较,突出承诺模型与其它协作模型相比的优势;最后,介绍了异构Agent协作模型在智能足球机器人、协同决策以及虚拟训练等领域的具体应用.在Agent协作模型的比较与具体应用中,突出了承诺模型在异构Agent协作过程中所展现的较好的动态性、灵活性和反应性等特征,并提出了在开放环境下,当前Agent承诺模型在动态生成承诺协议方面的不足与改进方向.

To begin with, we summarized the multi-agent theories and collaborative mechanisms based on analyzing the background and current situation of agent collaboration research. Then, after the study of agent collaborative framework, we analyzed some most representative collaboration models, such as blackboard model, contract model, acquaintance model, relation web model and commitment model, illustrating and comparing their characteristics in the aspects of dynamic, heterogeneity, reaction and flexibility to reflect the advantages of commitment model. Finally, we introduced the application of agent collaborative models in the fields of artificial soccer robots, integrating decision and virtual training. By comparing different collaboration models and describing their applications, we presented the advantages of commitment model in the aspects of dynamics, flexibility and reaction and tried to improve the current agent commitment model and proposed the improvement direction of the model.