Building Intelligent DSS Based on Multi-Agent Cooperation

A general multi-agent architecture is proposed for intelligent decision support system(MAIDSS). Theagent in MAIDSS is built based on an extension of BDI framework. Several agents form a team working togetheron a decision problem; several agent teams are defined to stand for the benefits of different people in the real     

Agent Belief with Credibility and Its Revision Mechanism

An approach to characterize the credibility of beliefs of an agent is proposed in this paper, which can define the uncertainty of beliefs, calculation rules and inference rules about credibility and a method for belief revision based on abductive reasoning is also given. When an agent receives some new information, if the new information is consistent with the current belief set, then incorporate this new information with an appropriate credibility, otherwise the choice will be different depending on the characters of agents, and the deliberated agents will choose the belief with a better explanation under the current belief set. Removing one belief may cause the removal of those beliefs that, together with others, logically entail the formula to be removed. A method based on abduction is proposed to solve these problems.     

局部特征在航拍图像拼接中的应用

针对航拍图像的特点和难点, 提出了一种基于局部特征的航拍图像拼接方法, 以提高拼接的质量和速度. 该方法利用多分辨率技术和局部特征对航拍图像序列进行配准. 通过仿射变换模型将拼接问题转化为像素点空间坐标变换的过程. 采用渐入渐出的图像融合算法, 消除拼接痕迹, 实现了色彩和亮度的平滑过渡. 实验结果表明, 该方法对航拍图像序列和视频图像拼接效果良好, 具有较高的实用价值.     

基于溯因推理的主体协商

为了能够进行有效的协商，主体应当提高通信的效率。为此，接收者可以对发送者的当时的意识状态进行推测，这可以通过溯因推理实现。该文提出了一个基于溯因推理的主体协商模型，是对Parsons的基于论据的协商模型的改进。     

基于双向推理的主体框架

文章提出了一个基于双向推理的主体框架FBRA,它是一个混合型的主体框架,主体既是反应的又是慎思的。它的推理内核是正向推理和反向推理相结合。正向推理用于对环境的反应,包括对其他主体的反应。反向推理基于溯因推理,用于信念修正、规划、多主体协调和多主体通信等。     

基于全局覆盖机制与表示学习的生成式知识问答技术

针对现有生成式问答模型中陌生词汇导致答案准确率低下的问题和模式混乱导致的词汇重复问题, 本文提出引入知识表示学习结果的方法提高模型识别陌生词汇的能力, 提高模型准确率. 同时本文提出使用全局覆盖机制以平衡不同模式答案生成的概率, 减少由预测模式混乱导致的重复输出问题, 提高答案的质量. 本文在知识问答模型基础上结合知识表示学习的推理结果, 使模型具备模糊回答的能力. 在合成数据集和现实世界数据集上的实验证明了本模型能够有效地提高生成答案的质量, 能对推理知识进行模糊回答.