Integrated RFID Data Modeling: An Approach for Querying Physical Objects in Pervasive Computing Shaorong Liu∗
Fusheng Wang and Peiya Liu
IBM Silicon Valley Lab San Jose, California, USA
[email protected]
Integrated Data Systems Department Siemens Corporate Research Princeton, New Jersey, USA {fusheng.wang,peiya.liu}@siemens.com
Categories and Subject Descriptors H.2.1 [Database Management]: Logical Design—Data models, Schema and subschemea
General Terms
Tag Type Reader/Location/Operation Fixed Reader Moveable Reader
Class 0,1 Read-only
Class 3 Class 2 Sensor-write Reader-write (Semi-Passive)
Fixed Location
A
F
G
No Location
B
-
-
Discrete Location
C
-
-
Continuous Location
D
-
-
E
-
-
Management, Design
With Operation
Keywords
Figure 1: RFID application scenarios
RFID, sensors, EPC, RFID data modeling, pervasive computing
1.
INTRODUCTION
Radio frequency identification(RFID) technology uses radiofrequency waves to transfer data between readers and movable tagged objects without line of sight. RFID holds the promise of real-time identifying, locating, tracking and monitoring physical objects, and can be used for a wide range of pervasive computing applications. To achieve these goals, RFID data have to be collected, transformed and expressively modeled as their virtual counterparts in the virtual world. RFID data, however, have their own unique characteristics – including aggregation, location, temporal and history-oriented – which have to be fully considered and integrated into the data model. The diversity of RFID applications pose further challenges to a generalized framework for RFID data modeling. In this paper, we explore the fundamental characteristics of RFID data and classify RFID applications into a set of typical scenarios. We then propose a generalized RFID data modeling framework with constructs for each typical scenario. These constructs can be combined to model most RFID applications in real world.
2.
SPECTRUM OF RFID APPLICATIONS
Tags and readers are two key elements in RFID applications. RFID tags, uniquely identified by their Electronic Product Code(EPC) IDs stored in memories, range from read-only to read-write including both reader-write (writable by readers) and sensor-write (writable by sensors). Readers can be either fixed or mobile. Fixed readers are statically mounted and observe data from tags within their scopes. ∗
Work done while visiting Siemens Corporate Research.
Copyright is held by the author/owner(s). CIKM’06, November 5–11, 2006, Arlington, Virginia, USA. ACM 1-59593-433-2/06/0011.
Mobile readers, either operated by humans or guided by motion control systems, move around and approach tags to read data. For applications with mobile readers, one special case is that these readers may act as surrogates of operators and observations are used to track operations. In such applications, operation, representing a process or an aggregated event, is a fundamental concept. RFID applications track objects by tracing their movements among different locations through observations. The semantics of a location can be either geographic (e.g., a location from a GPS system) or symbolic (e.g., a shipping route). RFID observations signify the processes or movements of tagged objects in applications, e.g., the movement of products in a supply chain. Based on the diversity of tags, readers and location semantics, we classify RFID application scenarios as shown in Figure 1. Scenarios A-G are fundamental ones: RFID applications in real world are usually combinations of scenarios A-G. Scenario A represents applications with read-only tags and fixed readers. Scenarios B, C and D are all for applications with read-only tags and moveable reader, but with different location semantics. Scenario E describes applications in which a reader is attached to an operator and an observation signifies the start or end of an operation. Scenario F (G) is represents applications with reader-write (sensor-write) tags and fixed readers.
3.
A GENERAL TEMPORAL DATA MODEL FOR RFID APPLICATIONS: TMDR
In this section, we discuss how to model RFID applications as discussed in Section 2. We first identify the fundamental entities in these applications and study relationships among these entities. We then propose a general data model for RFID applications. While there can be many entities in RFID applications, only some of them are directly related to RFID, which we consider as fundamental entities in RFID applications. These
include EPC-tagged objects, readers, sensors, operators, locations and transactions. A sensor measures a target and then writes the measurement to its master RFID tag. Operators refer to humans who operate with readers. For example, a nurse wearing a wearable reader may interact with syringes, medicines and patients. The nurse is an operator, and the interactions between the operator and objects represent certain operations. While the entities are static in general, the relationships among these entities can be either static or dynamic. Static relationships are similar to those in the traditional ER model. For example, the relationship between an object and its onboard sensor – OBJECTSENSOR – is a static relationship. Most relationships in RFID applications, however, are dynamic and history-oriented due to the temporal nature of RFID data. Interactions among entities may generate movement, workflow, operations, and business logic. The interactions are in two forms: state changes and events. State changes include change of object locations, change of object aggregation relationships, start/end of an operation and change of reader locations. State change history, i.e., the information about during which period an object is in a certain state, is essential to tracking and monitoring applications and shall be captured in RFID data models. Events generated during entity interactions include: • Observations. These are generated when readers interact with tagged objects. • Sensor Measurements. These are generated when a sensor on a tag senses a target, such as the measurement of temperature. • Property Values. These are generated when a reader writes the value of an object property into the object’s tag. For example, a reader may write the processing steps performed on an object to the tag attached to the object. Here, processing steps are the properties. • Transacted items. These are generated when an object participates in a transaction. To model the entities and relationships discussed above, we propose a general data model for RFID applications (DMRA). This model extends ER model for modeling static entities and relationships with the following new features. • Temporal Relationships There are two types of temporal relationships among RFID entities: relationships that generate events and relationships that generate state histories. For an eventbased relationship, we use an attribute timestamp to represent the occurrence timestamp of the event. For a state-based temporal relationship, we use attributes tstart and tend to represent the lifespan of the state. • Nested Relations Nested relationship is a new characteristic in readwrite RFID applications. For example, for an application with sensor-write tags, an onboard sensor records the temperature measurement history in the tag. Thus a reader observation contains both the EPC of the tag and the measurement history, i.e., a nested relation.
4.
A DATA MODEL EXAMPLE
READER READERLOCATION timestamp OBSERVATION
LOCATION tstart
OBJECTLOCATION
tend
OBJECTSENSOR
OBJECT timestamp TRANSACTIONITEM
tstart
timestamp S-OBSERVATION
tend CONTAINMENT
TRANSACTION
State-based Dynamic Relationship
SENSOR timestamp SENSOR MEASUREMENT TARGET
Event-based Dynamic Relationship
Figure 2: Data model for application scenario G In this section, we present a sample data model (Figure 2) for applications with fixed readers and tags writable by onboard sensors, i.e., scenario G. Sensors detect targets independent of readers and periodically write sensor measurements to tags. When a reader interacts with an object, the reader observes both the EPC of the tag and the logged sensor measurement history. This model contains two state-based dynamic relations in this model: OBJECTLOCATION and CONTAINMENT. OBJECTLOCATION preserves the location history of each object: the period [tstart, tend] during which an object stays in a location. CONTAINMENT records in what period [tstart, tend] an object is contained in its parent object. Besides state-based relations, there are also four eventbased dynamic relations: OBSERVATION, SENSORMEASUREMENT, S-OBSERVATION and TRANSACTIONITEM. OBSERVATION records the raw reading data generated from readers and SENSORMEASURMENT records the measurement data generated from sensors. S-OBSERVATION represents a reader’s observation of an object and its logged sensor measurement history. TRANSACTIONITEM records the event generated during the interaction between a transaction and an item.
5.
RELATED WORK
RFID technology has emerged for years and poses new challenges for data management [1, 2]. Little research, however, has been conducted on how to effectively model RFID data. Harrison et al [3] summarized RFID data characteristics and provided some reference relations to model the data. In their model, RFID data are modeled as events. Thus, RFID state history and temporal semantics of business processes are implicit. A RFID data model was developed in [4], which focuses on fixed readers with read-only tags.
6.
REFERENCES
[1] S. S. Chawathe, V. Krishnamurthy, S. Ramachandrany, and S. Sarma. Managing RFID Data. In VLDB, 2004. [2] F. Wang, S. Liu, P. Liu, and Y. Bai. Bridging Physical and Virtual Worlds: Complex Event Processing for RFID Data Streams. In EDBT, 2006. [3] M. Harrison. EPC Information Service - Data Model and Queries. Technical report, Auto-ID Center, 2003. [4] F. Wang and P. Liu. Temporal Management of RFID Data. In VLDB, 2005.