Sep 09, 2002Sept. 9, 2002 - There has been a great deal of confusion in the marketplace surrounding the Auto-ID Center's quest to build an "Internet of things" – that is, the world's first open, global network for tracking goods using low-cost RFID tags and readers. Many people say it will take years to create low-cost tags and develop the necessary infrastructure to track unique items globally. In this first installment of our special report Low-Cost RFID: The Way Forward, we will answer some of the key questions many companies have about the Auto-ID Center.
The center's goal is to create a global network that would make it possible for a company in China to tag a pair of sneakers or a polo shirt and have that item tracked as it moves through the supply chain to a retail outlet in the United States, Russia, Germany, or Argentina. The center is creating a layer of information technology that will sit on top of the existing Internet and provide a low-cost way of transmitting information about specific products.
Unlike most RFID systems, the Auto-ID Center's is based on the idea that each product will have simply an electronic product code (EPC) – a unique "license plate" that identifies the manufacturer, product category and unique item. All other information about that item – when it was made, where, how much it cost and so on -- will be stored in databases on the Internet or on a company's local area network.
The reason for using just a serial number on the tag is it makes the microchip much simpler and the cost of the tag much lower. There are other advantages, according to the Auto-ID Center, such as the ability to store a great deal more information about each individual item in a database. For instance, a robot on an assembly line might be equipped with a reader that enables the robot to identify parts. It would then be able to download instructions from an associated database on how to install different parts on different models. If that information were stored in the tag, the chip would be prohibitively expensive – which, of course, is why few people use such systems today.
Critics of this license plate approach say that the IT infrastructure needed to make the system work is years away or that it is just impractical. Often, this point of view is espoused by existing RFID tag makers, who would prefer to sell more sophisticated chips with higher margins. Other critics say companies will never trust their supply chain operations to the Internet because the Internet is unreliable. Not only are many Global 1000 companies moving in this direction, but companies like Amazon, which has more than a billion dollars a year in sales, exist entirely on the Web.
There will be times when storing more than just a license plate will be an advantage. In our case study The Art of Tracking Masterpieces, the museums felt the key to the system was the amount of data that could be stored on tags. Storing information about the condition of a painting on the tag enabled workers to retrieve that information with a handheld reader while the work of art was in transit. But looking four or five years out, it is likely that mobile data networks will become extremely common and data transfer rates will improve to the point where there may be little or no added benefit to storing data on a tag.
So is the infrastructure years away, as some claim? Yes and no. The Auto-ID Center's proposed system consists of four key elements needed to use the Internet to identify items. They are:
Electronic Product Code: - The EPC is a 96-bit number – the license plate – that will be stored on the tag. The Auto-ID Center's EPC is supported by the Uniform Code Council and EAN International, the two main bar code standards bodies.
The Object Name Service: - When a reader picks up an EPC, a computer will need to find information about the product. The computer will go to ONS to look up where a file on the product is stored on the Internet.
Physical Markup Language: - PML is an XML-based language that aims to describe products in a way computers can understand. PML will make it possible to search for all carbonated drinks in a warehouse, or all jeans.
Savants: - A Savant is a piece of distributed software that will gather data from an RFID reader and clean it. If two readers pick up the same EPC, for example, the Savant would eliminate the duplicate product code. The Savants are set up to pass along only relevant information, which prevents existing databases from being overloaded.
Each of these elements is already being used in the Auto-ID Center field test. Sponsors involved in the field test say all of the systems are functioning properly. The center is now focused on enhancing the infrastructure so that it can accommodate billions and eventually trillions of EPC reads, so that it has redundancies in case a server or section of the Internet goes down, and so that it can compensate for the limitations of the hardware. For instance, a reader may pick up a signal for a tag ten times, miss once because of interference and then pick it up again. Intelligent software needs to be developed analyze data and determine when a product was moved and when the reader simply failed to pick up the signal.
Much of the media focus on the Auto-ID Center has been on the quest for a 5-cent tag. In fact, the center won't produce any hardware. However, the center's position is that the infrastructure it has designed is worthless without low-cost tags and readers. So it has been working with its technology sponsors – principally Alien Technology and Matrics -- to develop a a low-cost tag that can carry the 96-bit EPC. Alien is making the tag cheaper by by making the chips extremely small.
The center also hired a startup engineering company called ThingMagic to design a specification for a low-cost reader that could be networked and could read tags at both 900 MHz and 13.56 MHz (see Conjuring Up a Low-Cost EPC Reader).
The single biggest technological challenge the Auto-ID Center faced was getting the chip specification right. Often, chip designs take several tries to perfect, and it can take six months or more for each new design to be turned into test chips. The center's ambitious schedule to have all the technology in place for its field test by the end of this year would have gone up in smoke if the readers couldn't read the tags. At a center board meeting in Cambridge in June, a ThingMagic prototype was able to read the EPC off of an Alien tag on the first try.
The Auto-ID Center has asked ThingMagic to produce 100 readers for the field test the center has been conducting in the United States since September 2001. Alien has delivered some 300,000 low-cost RFID tags that carry only EPCs. The tags and readers are not on the market, obviously, but they do exist, and they will be used in the third phase of Auto-ID Center's field test, starting in January, to track unique items.
The Auto-ID Center's system is a few years away from making it possible for robots to download instructions about how to assemble a car or for your clothes washer to know how to wash your l'espace cashmere sweater. Much of the work needed to fulfill the ultimate promise of the system will need to be done by professional programmers working for software companies, just as the Internet was designed in academia and made commercial by private companies. But if the Auto-ID Center's system is able to track several hundred thousand items during phase three of the field test, then clearly it will be far enough along to be used in warehouses and distribution centers to track pallets and cases of goods.
Obviously, having a working system doesn't mean that everyone is going to run out and adopt the Auto-ID Center's technology. It will only be adopted if companies believe it will provide a significant return on what will inevitably be a major investment. One of the keys to achieving a significant ROI will be not just the cost of the tags, but also the cost of readers, software and the IT infrastructure needed to track items.
The Auto-ID Center's goal was to create reference designs that would make it possible for private companies to produce EPC tags that cost about 5 cents and readers that cost about $100 when manufactured in significant volumes. This has been the most controversial area of the project.
Alien says that it will be able to produce RFID tags in bulk for less than 10 cents starting in the first quarter of 2003. As production ramps up, the price per tag is expected to fall to below 5 cents by the end of 2005. Established vendors selling tags for 50 cents or more scoff at this. However, insiders say companies like Coca-Cola are investing millions in Alien to ensure their supply of low-cost tags. And RFID Journal has learned that Alien is working on a strap that would cost just 2.5 cents. It won't have an antenna. Instead, it will use conductive polymers put in the ink used to print on a cereal box or other container. The strap could be on the market within three years.
The Auto-ID Center has issued white papers spelling out the business case for adopting its system. The papers use a base price of 40 cents per tag this year and anticipate that falling to 5 cents by 2006, as the number of tags sold rises from 200 million to 15 billion. The center says those estimates are conservative and our research supports that claim. Alien's strategic partners – customers that have invested in the startup – will likely pay no more than 10 to 15 cents for tags next year and certainly less than that by 2005.
Sources within the Auto-ID Center have told us that a number of technology companies that are sponsoring the Center have already committed resources to developing readers. NCR is working on a point-of-sale device that will read both bar codes and RFID tags. Symbol Technologies is working on a handheld unit that will read both bar codes and RFID tags. Neither company would confirm this information or provide a timeframe for when their readers might be ready.
What is known, however, is that ThingMagic is working on a shelf reader, in addition to a reader that can be installed on a dock door or on a forklift. The readers will be manufactured by a third party and will have two antenna ports for each of two frequencies -- 900 MHz and 13.56 KHz. The original aim was for the readers to cost $100 when manufactured in volume. ThingMagic's Bernd Schoner told us that the readers would initially cost from $500 to $1,000, in small volumes.
It is difficult to say how quickly the price of the readers will come down. The Auto-ID Center's white papers say that if 100,000 are produced this year, the readers will cost $500. As the number of readers sold rises to 2 million by 2006, the papers estimate that the cost of the readers will fall to $70. But will companies buy 2 million readers in 2006?
The center faces a common chicken-and-egg problem. Most companies won't adopt RFID technology until the cost of tags and readers come down. And the cost of the hardware won't come down until companies start deploying the technology. We believe the biggest hurdle the Auto-ID Center system faces is getting enough companies to use it initially. Think of the system as a giant snowball at the top of a mountain. Getting it moving isn't easy, but once it starts rolling down hill, it grows quickly and becomes unstoppable.