| 引用本文: | 谭开成,罗继亮,林鑫杰,章宏彬.基于知识Petri网的确定性和不确定性联合推理[J].控制理论与应用,2023,40(3):531~539.[点击复制] |
| TAN Kai-cheng,LUO Ji-liang,LIN Xin-jie,ZHANG Hong-bin.Certainty and uncertainty joint reasoning based on knowledge Petri nets[J].Control Theory and Technology,2023,40(3):531~539.[点击复制] |
|
| 基于知识Petri网的确定性和不确定性联合推理 |
| Certainty and uncertainty joint reasoning based on knowledge Petri nets |
| 摘要点击 1030 全文点击 353 投稿时间：2021-12-16 修订日期：2023-02-18 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2022.11234 |
| 2023,40(3):531-539 |
| 中文关键词 知识Petri网 确定性和不确定性 推理 剪枝处理 wumpus世界 人工智能 |
| 英文关键词 knowledge Petri nets certainty and uncertainty reasoning pruning wumpus world artificial intelligence |
| 基金项目 国家自然科学基金项目(61973130)资助. |
|
| 中文摘要 |
| 知识推理是人工智能的核心领域, 旨在研究如何从已知(知识库和推理规则)推理出未知, 以帮助智能体做
出科学决策. 而智能体所处的环境存在不可观性和不确定性, 因此知识库通常不仅包含确定性知识, 还包含不确定
性知识, 而且推理过程需要两类知识紧密协作. 然而, 目前的推理方法无法将两类知识统一表示, 常常将两者对应的
推理过程割裂进行. 基于此, 为了实现在统一的模型架构下完成确定性和不确定性联合推理, 给出了一种知识Petri
网推理方法. 首先, 定义了一种新的知识Petri网, 使其不仅能够描述确定性的知识规范, 也可以描述先验概率知识;
其次, 根据知识Petri网的网结构, 给出了一种知识Petri网概率独立剪枝算法, 能够指数级地降低不确定性推理的计
算复杂性; 最后, 利用知识Petri网及其概率独立剪枝算法, 给出了一种新型推理算法, 实现了确定性和不确定性的联
合推理, 并利用Wumpus 世界进行了演示和验证. |
| 英文摘要 |
| The knowledge reasoning is a core area of artificial intelligence, which aims to investigate how to reason
from known (knowledge bases and inference rules) to unknown in order to assist an agent to make rational decisions.
Since an environment where an agent lives may be unobservable and uncertain, its knowledge base usually contains both
deterministic and uncertain rules, and the reasoning process requires their close collaboration. However, they cannot be
represented in a unified way by the reported methods, and their corresponding reasoning processes are often separated
from each other. Therefore, an inference method is proposed based on a knowledge Petri net to realize the certainty and
uncertainty joint reasoning in a unified architecture. First, a new knowledge Petri net is defined, which can be used to
describe not only deterministic rules but also prior probabilities. Second, according to structures of a knowledge Petri net, a
probabilistic independent pruning algorithm is given, which can greatly reduce the computational complexity of uncertainty
reasoning. Finally, a new inference algorithm is given to realize the joint inference by utilizing a knowledge Petri net and
the probabilistic independent pruning algorithm, and the Wumpus world is taken as an example to illustrate and verify the
theoretic results. |
|
|
|
|
|