智能家居场景联动中基于知识图谱的隐式冲突检测方法研究  

Research on Implicit Interference Detection Based on Knowledge Graph in Smart Home Automation

作　　者：肖丁 王乾宇 蔡铭[1] 李秀 XIAO Ding;WANG Qian-Yu;CAI Ming;LI Xiu(College of Computer Science and Technology,Zhejiang University,Hangzhou 310027)

机构地区：[1]浙江大学计算机科学与技术学院

出　　处：《计算机学报》2019年第6期1190-1204,共15页Chinese Journal of Computers

基　　金：国家自然科学基金(51775496,51875503)资助~~

摘　　要：智能家居场景联动是智能家居系统提供的一种自动化服务,它通过各类智能设备的互联、互通和互动,从而为住户提供更为舒适、安全和节能的家居环境.然而,由于各个智能场景的侧重点不同、管理策略各异,当多个智能场景联动和叠加时,会引起作动器设备之间的竞争、干扰和矛盾等控制冲突现象,降低智能场景用户体验,缩短智能设备使用寿命,甚至危及住户的人身财产安全.因此,智能家居场景联动中的控制冲突问题亟需大量实验研究并找到解决方案.本文中,我们将控制冲突分为两大类,一类是同一作动器上执行明显相反动作的显式冲突,一类是不同作动器之间产生相互干扰动作的隐式冲突.现有的控制冲突检测技术对于显式冲突检测的研究较为成熟,而在隐式冲突检测方面性能相对较弱,其原因可归结为自动化程度低与漏判率高两大缺点.本文提出了一种基于知识图谱语义分析的作动器隐式冲突检测方法,可以充分整合并有效挖掘各类作动器功能描述的常识性知识,进而实现对作动器隐式冲突的自动化检测.该方法有两个主要环节,分别是作动器设备注册环节与规则制定环节.我们在作动器设备注册环节提取知识图谱中的常识性知识,对作动器功能之间的效果冲突进行识别判断,并以隐式冲突矩阵的形式将判断结果在边缘设备上进行存储,实现存在于不同类型作动器之间的干扰检测;在规则制定环节,借助隐式冲突矩阵的冲突信息进行字典式查询与匹配,以判断规则之间的隐式冲突.此外,基于知识图谱中设备功能短语自身特点,设计了基于语义贡献度的权重分配策略,完成词向量到短语向量的转化.为解决聚类后相同簇中近反义功能短语混合杂糅的问题,本文提出了相对极性指标和近反义关系一步扩展方法,并借由图模型染色算法将反义关系的短语拆分到不同子�Smart home automation is a service which provides residents with more comfortable, approved and energy-saving home environments via the connection, interoperation and interaction of smart devices. However, due to different concerns and management strategies of each smart scene, interferences such as competition, interference and contradiction among devices frequently occur when multiple automation scenes encounter. They influence user experience of smart home automation, shorten service life of smart devices and even endanger the personal and property safety of residents. Therefore, interference problems in smart home automation require enough researches and solutions urgently. In this paper, we divide the interference into two categories: one is the explicit interference that performs as obviously opposite actions on the same actuator, the other is the implicit interference that manifests as interference between different actuators. The state - of-the - art interference detection technology is relatively mature on explicit interferences. While in respect to implicit interferences, the performance of these detection technology is relatively weak, and the reasons of that can be concluded as two major disadvantages, low degree of automation and high false negative rate. This paper proposes an implicit interference detection method. Through analyzing information from knowledge graph, our method is able to integrate and utilize commonsense knowledge of actuators in different categories substantially, and then detect the implicit interference among actuators in automation. Our method contains two main phases, which are actuator registration phase and rule subscription phase. During the actuator registration phase, our method utilizes commonsense knowledge included in the knowledge graph to evaluate the effects caused by the actuator, judge the interference of effects automatically and at last store the results as the implicit interference matrix at the edge devices;While in the rule subscription phase, our method matc

关 键 词：智能家居场景联动 控制冲突 隐式冲突 知识图谱 权重分配 极性分析 智能家居 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]