Study of CEP-Based RFID Data Processing Model
Study of CEP-Based RFID Data Processing Model
1. Introduction
1.1. Background
RFID Radio Frequency IDentification as a new real-time monitoring technology has its own characteristics that differs from other real-time monitoring technology. RFID can collect real-time information and detect real-time event automatically. However, this technology presents the following challenges: 1) RFID reader could gather large quantity of primitive data that causes numerous primitive events.
978-0-7695-3497-8/08 $25.00 2008 IEEE DOI 10.1109/IITA.2008.213 254
Authorized licensed use limited to: Feng Chia University. Downloaded on January 9, 2009 at 04:30 from IEEE Xplore. Restrictions apply.
event (reader ID, tag ID, Tstart, Tend, interval) where reader ID represents the reader which exists in the event, tag ID represents the tag which exists in the event, Tstart represents the start time of the event, Tend represents the end time of the event, interval represents the continuous time of the event.
4. BCPS architecture
255
Authorized licensed use limited to: Feng Chia University. Downloaded on January 9, 2009 at 04:30 from IEEE Xplore. Restrictions apply.
The architecture of BCEPS system is illustrated in Figure 2. The entire system consists of five layers. The bottom layer is Device Layer that contains RFID readers which detect RFID tags and gather RFID data. The layer above Device Layer is Data Stream Processing Layer which is responsible for processing primitive RFID data according to some RFID rules and generating primitive events. The primitive events are then fed to Event Processing Engine that is the core of the BCEPS system. Event Processing Engine uses event constructor to produce complex events and then Event Processing System processes event stream to deliver the results into RFID Database which is based on PostgreSQL Database in Data Storage Layer. Users can use computer to deliver GUI Event to Event Processing Engine and then get the results from Event Processing Engine. The following are the details about the five layers.
processing copes with idiosyncrasies of readers and performs data cleaning, such as filtering and smoothing. This is important as RFID readings are known to be inaccurate and lossy. Some technologies are available in [5]. After data are processed, Data Stream Processing Layer will generate primitive event in light of the data and hand in the event to Event Processing Engine. Data Stream Processing Layer consists of the following steps, as illustrated in Figure 3.
(1) Anomaly Filtering Layer It removes spurious readings and readings that contain truncated ids. (2) Duplication Filtering Layer Data Stream Processing Layer gets primitive data from TDC (Temporal Data Cache, in Figure 4) which is like a queue that pop data from top and push data into bottom. Device Layer may generate duplicate primitive data which have the same reader ID, tag ID and only different timestamp. So the primitive data which has the largest timestamp will be saved and other duplicate primitive data will be discarded.
(3) Event Generation Layer It generate primitive event in light of a pre-defined schema. It retrieves attributes (employee no, reader no, label no, timestamp) defined in the schema from RFID database.
Authorized licensed use limited to: Feng Chia University. Downloaded on January 9, 2009 at 04:30 from IEEE Xplore. Restrictions apply.
Application. In order to meet the real time requirement of RTLSUM, we express any events in the format: events (type t, event e, priority p), where t represents the type of event which contains type of GUI events and type of primitive events, e represents the event itself and p represents the priority of the event. The larger the value of p is, the higher the priority of the event is. After event constructor finishes, the events generated are fed to Event Processing System. Event Processing System rearranges the events according to the priority of events. The event having higher priority will be processed and return results earlier. For example, there are two events occur: a GUI event A which wants to get the list of miners who are existing underground and a primitive event B which has real time location information and requires to update location information immediately. Event A is expressed in the format of events (GUI, A, pA), Event B is expressed in the format of events (primitive, B, pB), pB > pA. So Event Processing System will process Event B firstly, and then process Event A. This technology will ensure that the real time of the system is higher.
executing one of the events which have the highest priority 10.
Although real time monitoring is also a real time event, the priority of real time detecting is higher than real time monitoring so the event of real time detecting will be processed firstly. This events scheduling method arrange event executions more legitimately.
6. Conclusion
In this paper, we design and develop a prototype of BCEPS for processing RFID data stream. In order to enhance real time performance of system, we use a priority-based event scheduling method in the Event Processing Engine of BCEPS to ensure that real time event will be processed earlier. Finally we apply the BCEPS to our RFID-based Tracking and Locating System Underground Mine. Priority-based event scheduling method is a relatively simple algorithm. In the future, we will design an more effective event scheduling algorithm to improve real time performance.
Acknowledgement
This work is supported by the doctoral fund of the Ministry of Education (20070141037).
5. Experiment
Our RTLSUM makes use of the BCEPS model to enhance RFID data stream processing and improve real time of the system. Figure 5 contains three screenshots. The left and top screenshot represents a real time monitoring component that show the real time path of miners underground mine, the left and bottom screenshot represents a real time detecting component that gathers RFID data and the right screenshot represents Event Processing Engine. We can observe that Event Constructor receives 7 events and Event Processing System rearranges the 7 events in light of priority and is
References
[1] Jin Xingyi, Lee Xiaodong, Kong Ning, Yan Baoping (2008). Efficient Complex Event Processing over RFID Data Stream. Seventh IEEE/ACIS International Conference on Computer and Information Science. [2] E. Masciari (2007). RFID data management for effective objects tracking. the 23rd Annu. ACM Symposium on Applied Computing, 2007, pp.457-461.
257
Authorized licensed use limited to: Feng Chia University. Downloaded on January 9, 2009 at 04:30 from IEEE Xplore. Restrictions apply.
[3] Daniel Gyllstrom, Eugene Wu, Hee-Jin Chae, Yanlei Diao, Patrick Stahlberg, Gordon Anderson (2007). SASE: Complex Event Processing over Streams,. CIDR, Asilomar, CA, 2007. [4] B. Yijian, T. Hetal, L. Chang, W. Haixun and Z. Carlo (2006). A Data Stream Language and System Designed for Power and Extensibility. the 15th ACM Conf. on Information and Knowledge Management, Virginia, 2006, pp.337346. [5] SHENG WANG, SHAORONG LIU, PEIYA LIU, YIJIAN BAI(2006). Bridging Physical and Virtual Worlds: Complex Event Processing for RFID Data Streams. the 10th Int. Conf. on Extending Database Technology, Munich, Germany, 2006, pp.588607. [6] E. Wu, Y. Diao and S. Rizvi (2006), High Performance Complex Event Processing over Streams. SIGMOD, Chicago, 2006, pp.407418 [7] MIT Auto-ID Lab. EPC network architecture. January 2006. http://autoid.mit.edu/CS/files/3/networkarchitecture. [8] F. WANG, P. LIU (2005). Temporal management of RFID data. the 31st VLDB Conf., New York, 2005, pp.11281139. [9] Zimmer, D. and Unland R (1999), On the semantics of complex events in active database management systems. ICDE, 392-399, 1999. [10] S. CHAKRAVARTHY, V. KRISHNAPRASAD and E. ANWAR (1994), Composite events for active databases: Semantics contexts and detection. the 20th VLDB Conference, Santiago Chile, 1994, pp.10111205.
258
Authorized licensed use limited to: Feng Chia University. Downloaded on January 9, 2009 at 04:30 from IEEE Xplore. Restrictions apply.