Reordering of Location Identifiers for Indexing an RFID Tag Object Database

Radio frequency identification (RFID) has become one of the emerging technologies for a wide area of applications such as automated manufacturing, inventory tracking, and supply chain management. RF technologies make it possible to identify individual items in realtime by means of automatic and fast identification. Besides the real-time identification, RF technologies give additional advantages for monitoring of field-based operation by tracking and tracing the location of tags attached to items. By using queries on trajectories of RFID tag data, RFID applications can get events about field-based situation and then respond to them. To store and retrieve tag data efficiently, it is important to provide an index for the repository of tag data. The EPCglobal, being in charge of a standards management and development for RFID related technologies, proposes EPC Information Service (EPCIS) as the repository for tag events. The EPCIS is a standard interface for access and persistent storage of tag information. Tag data stored in the EPCIS consists of the static attribute data and the timestamped historical data. Historical information is continuously collected and updated whenever each tag is identified by an RFID reader. The EPCIS usually stores them at the base table of a database for efficient management of those data. It is necessary to execute queries on the EPCIS whenever applications want to retrieve the location history of specific tags. However, it is inefficient to look up all the records of the table because a large amount of historical information for tags is to be accumulated in the base table. For efficient query processing of tracing tags, an index structure can be constructed based on tag events generated when a tag goes in and out a location where a reader places. Among timestamped historical information contained in tag events, an RFID application uses the location identifier (LID), the tag identifier (TID), and the identified time (TIME) as predicates for tracking and tracing tags. To index those values efficiently, we can define the tag interval by means of two tag events generated when the tag enters and leaves a specific location, respectively. The tag interval could be represented and indexed as a timeparameterized line segment in a three-dimensional domain which is constituted by LID, TID, and TIME axes. Tag intervals in a three-dimensional index are sequentially stored and accessed in onedimensional disk storage. Since logically adjacent tag intervals are to be retrieved together at a query, they should not be stored far away from each other in the disk to minimize the cost