论文部分内容阅读
针对多Agent系统(MAS)中信任关系管理的需求,将Sarsa-强化学习(SRL)理论应用于构建MAS中基于Agent行为的信任关系预测模型.首先根据Agent之间交互的时间顺序,构建了基于时间戳的行为状态空间结构,然后应用SRL理论,建立了基于直接可信度和反馈可信度相融合的总体信任关系预测模型.新模型充分利用SRL理论较强的动态适应能力,解决了传统预测模型对环境的动态变化适应能力不足的问题.累计误差方面的实验结果表明,与已有模型相比,新模型能显著提高信任决策的准确性.
In view of the need of trust relationship management in MAS, Sarsa-Reinforcement Learning (SRL) theory is applied to the construction of trust behavior prediction model based on Agent behavior in MAS.Firstly, based on the time sequence of interaction between agents, Timestamp behavioral state space structure, and then apply the SRL theory to establish the overall trust relationship prediction model based on direct credibility and feedback credibility.The new model makes full use of the dynamic adaptability of SRL theory to solve the traditional The predictive model adapts to the dynamic changes of the environment.Experimental results on cumulative error show that the new model can significantly improve the accuracy of the trust decision compared with the existing models.