Tom Igoe - Getting Started with RFID: Identify Objects in the Physical World with Arduino

Beijing Cambridge Farnham Kln Sebastopol Tokyo

The process of identifying physical objects is such a fundamental part of our experience that we seldom think about how we do it. We use our senses, of course: we look at, feel, pick up, shake and listen to, smell, and taste objects until we have a reference for themthen we give them a label. The whole process relies on some pretty sophisticated work by our brains and bodies, and anyone whos ever dabbled in computer vision or artificial intelligence in general can tell you that teaching a computer to recognize physical objects is no small feat. Just as its easier to determine location by having a human being narrow it down for you, its easier to distinguish objects computationally if you can limit the field, and if you can label the important objects.

Just as we identify things using information from our senses, so do computers. They can identify physical objects only by using information from their sensors. One of the best-known digital identification techniques is radio frequency identification , or RFID. The network identity of a physical object can be centrally assigned and universally available, or it can be provisional. It can be used only by a small subset of devices on a larger network or used only for a short time. RFID is an interesting case in point. The RFID tag pasted on the side of a book may seem like a universal marker, but what it means depends on who reads it. The owner of a store may assign that tags number a place in his inventory, but to the consumer who buys it, it means nothing unless she has a tool to read it and a database in which to categorize it. She has no way of knowing what the number meant to the store owner unless she has access to his database. Perhaps he linked that ID tag number to the books title or to the date on which it arrived in the store. Once it leaves the store, he may delete it from his database, so it loses all meaning to him. The consumer, on the other hand, may link it to entirely different data in her own database, or she may choose to ignore it entirely, relying on other means to identify it. In other words, there is no central database linking RFID tags and the things theyre attached to, or whos possessed them.

Like locations, identities become more uniquely descriptive as the context they describe becomes larger. For example, knowing that my name is Tom doesnt give you much to go on. Knowing my last name narrows it down some more, but how effective that is depends on where youre looking. In the United States, there are dozens of Tom Igoes. In New York, there are at least three. When you need a unique identifier, you might choose a universal label, like using my Social Security number, or you might choose a provisional label, like calling me Franks son Tom. Which you choose depends on your needs in a given situation. Likewise, you may choose to identify physical objects on a network using universal identifiers, or you might choose to use provisional labels in a given temporary situation.

The capabilities assigned to an identifier can be fluid as well. Taking the RFID example again: in the store, a given tags number might be enough to set off alarms at the entrance gates or to cause a cash register to add a price to your total purchase. In another store, that same tag might be assigned no capabilities at all, even if its using the same protocol as other tags in the store. Confusion can set in when different contexts use similar identifiers. Have you ever left a store with a purchase and tripped the alarm, only to be waved on by the clerk who forgot to deactivate the tag on your purchase? Try walking into a Barnes & Noble bookstore with jeans you just bought at a Gap store, and you might trip the alarms if the two companies use the same RFID tags but dont set their security systems to filter out tags that are not in their inventory.