RFID Technology
Faculty Advisor, Umakishore Ramachandra, PhD
RFID Technology is rapidly becoming part of today’s life through a variety of applications in smart homes, airports, and supply chain management. With the ubiquity of RFID deployments, there exists a growing amount of data being generated from multiple sources that needs to be processed in order to support user queries. However, handling large volumes of sensor generated data streams stresses even high performance computing resources and broadband network bandwidth. It imposes several challenges, especially when these streams are distributed across multiple geographic locations. First, the amount of data from different sources that needs to be organized properly is very large; second, because of the nature of RFID applications, the streamed data and possibly legacy data must be fused together in a time-sensitive manner (e.g. using timestamps); and finally, the overall system has to be able to manage complexities in a scalable manner to answer queries efficiently. These problems become more severe in RFID systems where data is generated from error prone devices like RFID readers [9][8]. RFID readers use radio waves to communicate with electronic tags which vary in type, capability and size. Passive tags, in particular, are gaining considerable amount of attention as a low cost solution for automating a wide range of applications. However, one of the major drawbacks of the readers using passive tags is their unreliable behavior. Typical readers accurately detect tags 80-90% of the time, whereas readers with improved performance have a detection rate of 95-99%. Unfortunately, this rate is adversely affected by different environmental factors such as the presence of metal objects, objects containing liquids, interference from multiple readers, or the presence of multiple tags within close proximity of one another. For example, the detection rate drops to 70% when a reader attempts to detect more than five tags [5][8][9]. There have been several recent proposals involving middleware systems for RFID deployment. Savant [7] is designed to handle large amounts of data in a hierarchical manner by constructing a tree of subsystems. Similarly, the architecture defined in High Fan-in System [6] uses a tree like structure for data and query propagation. The problem is handled using a publish-subscribe model in the work of RFIDStack [4]. WinRFID [5] is another architecture that uses web services to ensure data availability. All of these systems aim to provide a scalable solution in distributed environments for filtering, organizing, and providing query results to the end user. The focus of our work is complementary to these systems. In our work, we focus on the inherent unreliability of RFID systems, and ask whether the reliability can be improved using a middleware system. We use the nature of the dataflow as our ally to improve reliability. Further, the data flow can help in data organization as well, which has been focus of earlier systems. For example, consider a typical RFID deployment scenario in a warehouse. RFID readers are placed at appropriate locations, and they continuously the read tags associated with the moving items, e.g. palettes. Often, palettes of a particular item type follow the same physical path from an entry point to the designated destination in the warehouse. Therefore, for many applications, generated tag data can be categorized based on their physical flow. Most RFID middleware support two major application categories: object tracking and object location. Both applications have an internal data flow that follows a path. In a typical supply chain scenario, as an example of an object tracking application[1], readers are placed along conveyor belts for item detection. Airport baggage claim systems are another example of object tracking applications [2]. In both these scenarios, readers are statically placed, and the tagged items detected by the readers construct a path. Conversely, in the case of object location applications, tagged items are placed statically and the set of readers are moving along a path locating items in their wake. Robots equipped with RFID readers looking for lost objects in an environment is an example of object location. Another scenario is using robots in a disaster situation: tags deployed in the environment may guide the robot in its recovery operation. As should be evident, a path naturally gets created following the sequence of tags acknowledged by readers. We have focused on the object tracking scenario for the work presented in this paper1. A middleware that has inherent support for this flow of data considered as a path can be very useful for organizing, filtering and thus responding faster to queries. Using the intuition of data flow, we have designed a path based distributed system architecture for RFID middleware called RF2ID (Reliable Framework for Radio Frequency Identification). Our proposed scheme consists of (1) a virtual reader abstraction to capture the static and potentially error-prone nature of the physical readers and antennas in a scalable manner; and (2) a novel path abstraction, called Vpath (Virtual Path) to capture the logical flow of 1 However, it should be noted that the architecture is general and applies to object location scenario as well. information among the virtual readers as RFID-tagged objects move throughout the environment. Using a notion of path at the system level gives us several advantages. First, the system load can be distributed among multiple virtual readers that constitute a specific virtual path. Second, different QoS attributes can be defined for a path, such as accuracy and priority levels that the virtual readers use to operate on data flowing through the corresponding path. Finally, as there is an internal representation of data based on path attributes, it becomes trivial to support path-related operations on the data, e.g., searching for query results, or making a future projection of data behavior based on history. We have conducted two set of experiments to evaluate RF2ID. The first sets of experiments evaluate the improved system performance in terms of reliability while the RFID resources are inherently unreliable. We have done an extensive study on the ALR-9800 [3] RFID readers. We studied tag behaviors using 6 passive RFID tags with reader power set to its maximum level (31.5 dB). The experiments show the key factors that affect the number of detected tags by a reader: varying the reader to tag distance, reader to tag angle and RF attenuation level. Lastly, we study how the tag reading varies over time when the above parameters remain unchanged. Then we explore the strength of the path based system to locate missing or misplaced items along its traversal. We showed by our experimental results using RFID readers and simulated readers, how our path based system improves reliability along with the capability to detect missing or misplaced items in the system.
References:
[1]. R. Weinstein, "RFID: A Technical Overview and its Application to the Enterprise," IT Professional, vol 7, 2005.
]. M. C. O'Connor, "McCarran Airport RFID System Takes Off," RFID Journal, http://www.rfidjournal.com/article/articleview/1949/1/1.
[3]. "Alien Technology RFID Readers", http://www.alientechnology.com/products/alr9800.php [4]. C. Floerkemeier et al., "RFID Middleware Design - Addressing Application Requirements and RFID," In Proceedings of sOc-EUSAI 2005.
[5]. B. S. Prabhu et al., "WinRFID – A Middleware for the enablement of Radio Frequency Identification. (RFID) based Applications," UCLA – Wireless Internet for the Mobile Enterprise Consortium.
[6]. M. J. Franklin et al, "Design considerations for High Fan-in Systems: The HiFi Approach," Proceedings of the 2nd CIDR Conference, 2005.
[7]. Oat Systems and MIT Auto-ID Center, "The Savant," Technical Manual. February, 2002.
[8]. "The Basics of RFID Technology," http://www.rfidjournal.com/article/articleview/1337/1/129/ [9]. T. Hassan et al.,"A Taxonomy for RFID," Proceedings of the 39th Annual Hawaii International Conference on System Sciences (HICSS'06)