Algorithms And System Implementation Of Distributed RFID Composite Event Detection

In recent years, RFID (Radio Frequency Identification) technology has posed many new challenges for data processing and management, arising from the problems caused by missed and unreliable RFID readings, redundant RFID data, in-flood of RFID data, and their spatial and temporal management. Complex Event Processing (CEP) is such a technology for low-latency filtering, correlating, aggregating, and computing on RFID data produced by large-scale applications. However, most of the existing CEP techniques adopt a centralized system architecture, which requires more bandwith and computational power for collecting real-time RFID data and processing the incoming data streams without latency.Since large-scale RFID applications are highly distributed, the focus of this thesis is the composite event detection in such a districuted computing environment. Motivated by current theories and algorithms in composite event detection, we design and implement a distributed RFID composite event detection system, which has a Master-Worker style parallelism for low-level data stream processing and can digest the data close to their source by using tree-based event detection algorithms. To do so, users specify their composite event expressions according to their requirements, RFID event processing system generates event matching tree model corresponding to user-defined expressions, next the system processes raw RFID raw data by using such a model, data structures generated previously, the processing result is finally sent to users.In this thesis, our implementation considers:(1) distributed architecture design, (2) the study of the specification of RFID composite event expression, (3) parsing of user-defined RFID composite event expression, (4) generation of event matching tree, (5) generation of RFID composite event filtering rules, and (6) processing of RFID composite event constructors. Finally we conclude and put forward the future work.