In RFID/USN (Radio Frequency IDentification/Ubiquitous Sensor Network) environments, a large amount of data is generated. Since data generated in these environments has different characteristics from data generated in traditional environments, we cannot apply traditional database technologies. We propose techniques to manage such data effectively and efficiently. We first discuss data management techniques in RFID environments and then discuss those in USN environments.
Since data generated in RFID environments has noisy and duplicate data, various filtering techniques are needed. Especially, it is important to eliminate duplicate data since RFID data has many duplicates. However, it is difficult to eliminate duplicate RFID data in one pass with the limited memory since a large amount of RFID data is generated simultaneously in a streaming fashion. Therefore, we propose one pass approximate methods (Time Bloom Filters and Time Interval Bloom Filters) based on Bloom Filters using a small amount of memory.
The RFID data collected from many RFID readers will be used for various analyses in the central server. It is important to analyze the object transition in many RFID applications. In this dissertation, focusing on supply chain management, we propose an efficient storage scheme and query processing for large RFID data sets in order to analyze the object transition efficiently. We first define query templates to analyze the object transition. We then propose an effective path encoding scheme to encode the flow information for products. To retrieve the time information for products efficiently, we utilize a numbering scheme used in the XML area.
The generated data in USN environments is ordered data and its volume is huge. Therefore, we need to reduce the ordered data effectively. For effective data reduction, we consider compression schemes with data reordering. We first investigate general principles to improve the compression ratio for ordered data by reo...