中文摘要：

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．

英文摘要：

Focusing on the requirement of trust management in multi-Agent systems, the sarsa reinforcement learning （SRL） theory is applied to construct trust prediction model for multi-Agent systems based on Agent＇ s behavior. First, basic formal description is conducted for trust decision, and behavior state- space structure is constructed based on time-stamp according the interaction time sequence between network Agents. With SRL algorithm, overall trust relationship predicting model based on direct trust degree and feedback trust degree is proposed. The model makes full use of the advantages of the strong dynamic adaptive capacity of the SRL algorithm, brakes away from the inadequate dynamic adaptive capacity in the traditional software trust modeling process. Simulation in cumulative errors shows that, compared to the existing models the new model has remarkable enhancements in the trust decision accuracy.