RFID数据流上多目标复杂事件检测

Multiple Objects Event Detection over RFID Data Streams

已有的RFID复杂事件处理技术主要关注于单个RFID对象的复杂事件检测和优化技术.实际上,很多RFID应用中往往需要同时检测多个同类型关联目标的复杂事件序列.研究了多个关联的RFID对象的复杂事件处理问题.通过扩展的事件语言和算子的语义以支持同类型多个RFID目标复杂事件查询的定义.通过模式的变换规则,将RFID应用中存在的各种非线性多目标复杂事件模式转换成线性模式,以便各种多目标模式在一个统一的框架下检测.提出了基于自动机NFAb2的多目标复杂事件检测模型和多目标复杂事件检测算法.通过在多目标检测算法中使用关键节点下压和同位置约束置后优化策略,大大减少了单个类型上无用实例的数目和不同类型间模式匹配的搜索空间.与SASE算法的实验比较表明算法的正确性和高效性.

Complex event processing is a data analysis technology which is widely applied in timecritical applications such as l^FID-enabled object tracking, stock trend prediction and network intrusion detection, etc. In an RFID-enabled monitoring system, RFID objects are always tracked with complex event queries. Existing RFID complex event processing techniques mainly focus on event detection and optimizations over single RFID object. However, in many RFID scenarios （such as in an RFID-enabled office or auto assembly line）, complex events sequences of multiple co-located and correlated objects are always subscribed due to consistency checking and regularity requirements. In this paper, event processing issues over multip[e correlated RFID objects are investigated. To support multiple correlated objects event query definition, semantics of existing event operators is extended. With pattern transformation rules, non-linear multi-objects patterns are transformed into linear multi-objects patterns which can be well evaluated within a unified evaluation framework. A multiple co-related objects complex event query evaluation model called NFAbz and the corresponding event detection algorithms are proposed. By pushing check-point constraint check down and postponing co-location constraints in event detection process, unviable runtime instances and pattern matching search space are greatly reduced which can conserve huge CPU time. Empirical experimental analyses between the proposed multiple RFID objects event detection algorithm and the slightly altered popular event detection algorithm--SASE illustrate both the efficiency and soundness of the proposed algorithm.