RFID (Radio Frequency Identification) technology is enabling cost-effective sensing and identification in various large-scale applications in retail, pharmaceuticals, healthcare and supply chain management. However, the characteristics of RFID data include time-dependency, dynamic changes, huge volumes and hoard of implicit semantics. The inherent nature of RFID data poses significant challenges to RFID data management, problems become even worse when there are missed readings, false readings, and redundant readings.In recent years researchers have investigated ways of RFID data processing and integration, but there is still space for improvements in temporal data processing, modeling, and querying. In this thesis we propose spatio-temporal data models, design and implement a rule-based framework in support of efficient tracking and monitoring of RFID tagged objects. In doing this, we first present a comprehensive survey of the state-of-the-art in RFID data mangement. Then, we present an extended DRER model to centally handle massive RFID data in terms of filtering, aggregation, transformation, and spatial/temporal data management. To perform efficient temporal queries to track and monitor moving objects, we utilize in-memory caching, database partition and index techniques to optimize the proposed algorithms. The effectiveness of our algorithms is validated by extensive experiments that are conducted on real RFID data and simulated data. Furthermore we propose a distributed RFID data model by incorporating multiple DRER models with distributed indexing through GateNodes, which utilizes the arithmetical fundamental theorem to index moving objects in distributed sites. Also we have implemented a temporal RFID data management system by using the proposed algorithms. |