Article Preview
Top1. Introduction
With the rapid development and growing maturity in RFID (Radio Frequency Identification) technology in recent years, a large number of RFID devices have been widely used in many areas of modern life, such as supply chain management (Lee et al., 2008), health care (Kwon et al., 2013), air monitoring (Honda et al., 2014), power network management (Sitthidet et al., 2014), pervasive computing applications (Boytsov et al., 2013) and industrial manufacturing(Son et al., 2014). The wide usage of RFID devices generates massive RFID event stream. However, the characteristics of massive RFID event stream: huge volume, high velocity, many variety, small value and not veracity, etc., makes the existing event processing methods on event stream difficult to rapidly deal with the massive RFID event stream above, thus seriously affecting its whole processing efficiency. So, how to develop a technology to quickly get desired information from massive RFID event stream has become an important issue in this paper.
Since CEP (Complex Event Processing) (Del et al., 2007) technology can rapidly pick up valuable information from massive RFID event stream by taking advantage of the association between event attributes, matching rules and algebraic operations for users, therefore, it has obtained increasing attention in the field of event stream processing.
In recent, many researches on CEP have been done to detect a complex event from various streams. In the processing systems aspects, some general complex event processing systems, such as Cayuga (Demers et al., 2007), Esper (2010), Estream (Garg et al., 2005), PQS (Cybenko et al., 2007), have been developed to provide the basic processing functions for complex event. In processing methods aspects, some general complex event processing methods, for example, complex event processing method based on petri-nets (Wang et al., 2006), complex event processing method based on diagram (Bai et al., 2009), complex event processing method based on tree (Sun et al., 2009), complex event processing method based on finite automaton (Mei et al., 2009), complex event processing method based on workflow (Zang et al., 2009), etc., and their some optimized algorithms, such as complex event processing method based on timed petri-net (Jin et al., 2008), complex event processing method based on optimized directed graph (Wang et al., 2006), complex event processing method based on compressed composition tree (Li et al., 2010), complex event processing method based on pushdown automata structure (Cao et al., 20114), and so on, have been studied to detect a complex event. However, since these processing systems or methods above need to take many repeatable unnecessary inserting, storage and search operations when detecting a complex event for massive RFID events stream, which can result in long processing time, high memory consumption and low event throughput.