CAJ | 학술논문

为准确评价协作目标,在agent信任评价模型中引入动态学习机制.采用该机制,agent在协作中可以根据环境参数、历史交互信息动态调整信任评价方式,并且在模型中引入陌生人信誉,根据对陌生人熟悉度和可靠度的计算结果处理与陌生人的交互.实验表明,采用该机制,MAS在取得目标奖励的同时,可以动态选择信任评价方式,有效地对目标agent进行评价.
위준학평개협작목표,재agent신임평개모형중인입동태학습궤제.채용해궤제,agent재협작중가이근거배경삼수、역사교호신식동태조정신임평개방식,병차재모형중인입맥생인신예,근거대맥생인숙실도화가고도적계산결과처리여맥생인적교호.실험표명,채용해궤제,MAS재취득목표장려적동시,가이동태선택신임평개방식,유효지대목표agent진행평개.
This paper introduced the dynamic learning mechanism in the agent trust evaluation model to dynamically adjust the evaluation method, according to the environmental parameters and historical interactive information. Also proposed the stranger reputation in the model. The result of familiarity and credibility could manage the interaction with strangers. Simulation shows that the mechanism is able to dynamically select the evaluation method and effectively evaluate the object agent, and gain object award at the same time.