Issue No. 01 - January/February (2009 vol. 13)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MIC.2009.19
Sastry S. Duri , IBM T.J. Watson Research Center
George Roussos , University of London
Craig W. Thompson , University of Arkansas
After 60 years of development, RFID is finally reaching maturity and appears to be fulfilling its early promise as a stand-alone technology. The past few years have seen an explosion of interest in RFID across industry and academia and a variety of new applications.
Two main technological advances have made this explosion possible: the wider availability of very low-cost and higher–range passive RFID tags that require no battery to operate, and the use of the Internet to connect otherwise isolated RFID systems and software. The latter development, in particular, has allowed not only the integration of distinct RFID implementations but also, and more importantly, the development of critical network-based support services. Networked RFID is a significant breakthrough in recent times and addresses passive tags' limitations as regards communication, computation, and storage capacity. In this case, full system functionality is possible only through network support: rather than storing object-related data on a tag, tagged objects and their associated metadata can be matched through a task-specific network service.
RFID economics have also evolved. This is in part due to Moore's law and novel tag manufacturing processes, but targeted governmental and private–sector mandates have also fostered large-scale implementations in the supply chain, ticketing, asset tracking, maintenance, retail, and personal identification areas. These applications have established a critical mass, allowing RFID to grow rapidly and enabling further economies of scale. In turn, this has reignited interest in RFID for research because it offers unique opportunities for automatic identification over alternative wireless sensor networking and communication technologies embedded in physical environments. Today, many view RFID as a useful ingredient in building the so-called Internet of Things, a network-centric vision for pervasive computing. [A special IEEE Internet Computing track will explore certain aspects of this vision starting next issue. — eds.]
Complexity and the Internet
Networked RFID's complexity varies widely from application to application. At one end of the spectrum, local networking is enough, and Internet connectivity isn't required. For example, the Tread Act ( www.nhtsa.gov/cars/rules/rulings/index_treadact.html) demands that every car tire sold in the US carry an RFID tag, which holds all related descriptive and lifeline information. On the other hand, compliance with drug–pedigree mandates associates significant amounts of track-and-trace information with each drug item that far exceeds the storage capacity of any tag currently available on the market. Attempting to store such information in an RFID tag is neither practical nor advisable. Instead, each tag carries a simple object identifier that's employed as a handle linking the physical entity with information about it stored in the network.
RFID's reliance on the Internet to provide core functionality requires new supporting technologies. Various organizations have introduced new RFID-specific network services — often referred to as repository services — to support registration, mapping, and discovery of unique identifiers under various object-identification schemes. These include associated authentication services that support safe access to and updating of this information online, which is especially important when we take into account complex relationships between organizations and a particular object's individual custodians. Another area that's undergone considerable development is in-network processing techniques — especially at the edge of the network — that many have proposed in order to reduce RFID traffic forwarded toward the Internet's core. Such techniques include filtering, aggregation, and other data-shaping mechanisms and have been standardized in event-based middleware specifications, which currently provide the dominant programming models for network-based RFID. Last but not least, the RFID community has developed and tested new system architectures to address RFID's specific operational requirements. The International Standards Organization, the ITU, and the IETF are standardizing much of this work, as are several RFID-specific initiatives, including the uID Center and EPCglobal, the latter having established an early lead in deployments, especially in logistics.
These new RFID-related technologies are already deployed in operational systems, which are growing rapidly in size and popularity and are having an increasingly observable effect on Internet systems and applications. These systems' availability also helps identify interesting questions about future network developments: to a large extent, RFID applications reside at the network's edge and thus have architectural similarities with other emerging technologies, including P2P systems and sensor networks. In fact, significant overlap often occurs between these technologies because RFID systems can use additional wireless sensor network technologies and operate in P2P mode — for example, in the case of trading partners in a supply chain.
Although research in networked RFID has developed rapidly, it seems that practitioners could incorporate additional lessons learned from real-world applications involving planetary–scale services and content-distribution networks, for example Planetlab ( www.planet-lab.org), Ron ( http://nms.csail.mit.edu/ron/), and Opera ( www.cl.cam.ac.uk/research/srg/opera/). Applying such techniques to network RFID presents a clear opportunity for the networking research community. We must also consider data management issues, given that network RFID's particular characteristics can enable significant optimization and thus considerable performance gains. Exploring the spectrum of useful applications is another area that can provide interesting directions for future research, especially taking into account opportunities that modern computational intelligence and Web-based visualization techniques offer. Finally, the intimate linking between real and virtual entities that RFID enables creates major security and privacy risks we must manage in order to guarantee its safe use — work in securing tags themselves is well under way, but little effort has gone into security for networked RFID.
In this Issue
For this special issue, we've chosen three case studies on networked RFID that relate to middleware, services, overlaying, and the network edge. We believe that this set of topics will resonate with IEEE Internet Computing's readership, reflect this technology's implications, and possibly help identify potential areas of interest to the RFID community that haven't yet received appropriate attention.
In "RFID Infrastructure Design: A Case Study of Two Australian RFID Projects," John P.T. Mo, Quan Z. Sheng, Xue Li, and Sherali Zeadally report on lessons learned through their first-hand experience in designing, implementing, and deploying a large-scale RFID infrastructure. They discuss two national Australian RFID projects implemented on the EPCglobal Network architecture, arguably one of the most popular networked RFID specifications. This article also serves as a good introduction to EPCglobal standards' main features and elements, which should be valuable to readers not familiar with networked RFID infrastructures in general. The authors identify several areas of improvement research opportunities for such systems in the context of future large-scale RFID implementations. They also propose some solutions that they developed that perform well in the context of their work.
In "Transf-ID: Automatic ID and Data Capture for Rail Freight Asset Management," Jorge González Fernández, Juan C. Yelmo García, Yod-Samuel Martín García, and Jorge de Gracia Santos describe their experience designing and implementing an RFID-based tracking system for the rail freight industry. In addition to discussing how they built a system that brings information about locomotives and their mechanical parts to the Internet, they pay particular attention to middleware component development. Their overall aim is to design all elements of a networked RFID system in a way that provides an end-to-end solution to the problem of translating tag observation into meaningful application events at the business layer. To do this, they focus on recording information at specific checkpoints while locomotives are on the move, thus inferring valuable information about routes. This data improves maintenance task efficiency and lowers operating costs by ensuring that operators decide when to replace mechanical parts on the basis of mileage and wear, rather than just elapsed time since the parts' installation.
Finally, Alexander Ilic, Thomas Andersen, and Florian Michahelles present "Increasing Supply-Chain Visibility with Rule-Based RFID Data Analysis." They describe a prototype system used to increase the visibility of the flow of physical items and goods in supply chains to detect inefficiencies, such as counterfeiting, inventory shrinkage, and out-of-stock conditions. Their system collects supply-chain data through the EPCglobal Network, and they employ a mapping mashup to display significant observed events. Similar to Fernandez and his colleagues' system, this one also provides a higher-level view by aggregating numerous tag observations in a way that's application-appropriate.
Five years ago, networked RFID was still mostly experimental. Today, we're just passing beyond the early adopter phase as more and more industry sectors are using this technology for increasingly diverse applications. Direct experience indicates that RFID tagging at the container level provides return on investment. At the item level — for example, for higher-value apparel — retailers are discovering potential opportunities for extended customer service. We're also seeing the beginnings of RFID coupled with sensors, especially temperature sensors in the cold chain, the supply chain that transports food to market, which requires assurance that food arrives safely and is fresh.
If we look a little further out, we believe that RFID will soon enter the consumer era. Tags placed on individual product items will help inventory our possessions, capture related information (such as where and when we purchased an item), and even automate replenishment. This will make RFID integration into cell phones more common in the same way GPS is today. We expect this future RFID to be cheaper and smaller; it will certainly have to become much more secure. As the vanguard technology for a networked, sensor-rich world, RFID is also leading the way toward a smarter one.
George Roussos is a senior lecturer in the School of Computer Science and Information Systems at Birkbeck College, University of London. Roussos has a PhD in scientific computation from Imperial College London. Contact him at firstname.lastname@example.org.
Sastry S. Duri is a senior software engineer at IBM T.J. Watson Research Center. Duri has a PhD from the University of Illinois at Chicago. From 2004 to 2007, he represented IBM on the EPCglobal Filtering and Collection work group. Contact him at email@example.com.
Craig W. Thompson is a professor and the Charles Morgan Database Chair at the University of Arkansas. Thompson has a PhD in computer science from the University of Texas, Austin. He is an IEEE fellow. Contact him at firstname.lastname@example.org.