Identifying Trillions of Items
When we say we are creating an Internet of things, we aren't suggesting that we plan to build another global network. Rather, we plan to build on top of the Internet. We are developing only those elements that will enable RFID systems to track items and share information over the Internet. Still, making it possible to uniquely identify every item manufactured everywhere in the world is a huge and complex undertaking. Let's break down each aspect of the challenge and explain how we are tackling it.

How do you distinguish between one can of Coke and another?
There are a number of ways, but the best solution we've found is to give each item a unique number - a license plate, if you will. The Auto-ID Center has proposed a universal standard for product "license plates" - the Electronic Product Code. Like a bar code, the EPC is divided into numbers that identify the manufacturer, product, version and serial number. But the EPC uses an extra set of digits to identify unique items. The EPC is the only information stored on the RFID tag's microchip. This keeps the cost of the tag down and provides flexibility, since an infinite amount of dynamic data can be associated with the serial number in a database.

How do you track the item using the license plate?
The answer is to create a network of RFID readers (sometimes called interrogators). In a warehouse for example, there could be readers around the doors on a loading dock and on every bay. When a pallet of goods arrives, the reader on the dock door picks up its unique license plate. Inventory systems are alerted to its arrival. When the pallet is put in bay A, that reader sends a signal saying item 1-2345-67890 is in bay A.

How do you know what item 1-2345-67890 is?
The EPC by itself tells you no more about a product than a car's license plate tells you about a car. Computers need a way to associate the EPC with information stored somewhere else about the unique item. To help computer systems find and understand information about a product, the Auto-ID Center has developed some new technologies and standards. The first key element is called the Object Name Service. ONS points a computer to an address on the Internet where information about a product is stored. The concept is based on the Domain Name Service, which points computers to the address of particular Web sites on the World Wide Web. ONS basically tells a company's computer systems: "Everything you need to know about product 1-2345-67890 is stored in a file on a computer located at the following Internet address…"

How does a computer act on information about a product?
The point of automatic identification, of course, is to take people out of the loop, to enable computers to gather information and act on it. For that to happen, computers must be able to not just identify a product, but interpret some basic information about it. To make this possible, the Auto-ID Center has created a new computer language called the Physical Markup Language. PML is based on the widely accepted eXtensible Markup Language (XML), which is used to describe common types of data (addresses, dates, invoice numbers and so on) and transactions (purchases, requests for quotes and so on) in a way computers running different proprietary applications can understand. PML files will be stored in PML servers, dedicated computers that deliver information over a network. (The Object Name Service, described above, points computers to the PML server.)

Some information about each product will be stored in a PML file, such as a product’s name and broad category (soft drink, auto part, clothing and so on), when it was made and where, its expiration date, its current location, even its current temperature, if that’s important. PML files will provide information to existing enterprise applications or new yet-to-be developed applications. The PML file could contain instructions for where a pallet should be shipped. It could contain instructions for a point-of-sale display to lower the price of an item when its expiration date approaches. Or it could contain instructions for how long your microwave needs to cook a particular brand of frozen pizza.

How do you avoid having all this data about individual products overload existing networks?
The Auto-ID Center has created software technology called Savant to manage and move information in a way that doesn’t overload existing corporate and public networks. Savant uses a distributed architecture, meaning it runs on different computers distributed through an organization, rather than from one central computer. Savants are organized in a hierarchy and act as the nervous system of the new EPC network, managing the flow of information. At the edge of the network, Savants gather data from readers. They pass on only relevant information to existing business applications, such as which products are about to expire. A Savant running at a distribution center might determine when product needs to be reordered from manufacturers, and so on.

How do companies use the EPC data to become more efficient and more profitable?
How companies use EPC data and the network we are creating will be up to them, just as it’s up to them to decide how they want to use the Internet. But the Auto-ID Center is providing some basic tools that will help them take advantage of the network. Savant will have a Task Management System that will enable companies to set triggers. A Savant running in a manufacturing plant might send a message to a parts replenishment system indicating which station on an assembly line is running low. A Savant running in a store might signal an existing inventory system to delay a shipment of diapers because there are already too many pallets in stock.