中文摘要：

为了克服单个Agent知识的局限性，提高系统决策的可靠性，提出了一种基于证据推理和粒计算的Muhi．Agent决策信息融合算法，并对Muhi—Agent合作决策进行了定义和描述。Multi．Agent决策融合划分为观测和决策两个阶段，观测Agent从环境信息中提取特征向量作为输入，信息粒化后降低了合成计算的复杂度。

英文摘要：

In order to improve the reliability of the decision-making and overcome the knowledge limitation of a single agent, this paper introduced an information fusion model based on evidence reasoning and granular computing, and defined the cooperative deci- sion-making of Multi-Agent. The decision making fusion has been divided into two stages of observing and decision-making. The obser- vation Agent takes feature vectors as inputs from environment information, and the result is that the information granulation method in- troduced in this paper can reduce the complexity of synthetic computation effectively.