You’ve spent the last several weekends hunting for the perfect lamp to brighten up that shadowy corner of the living room. When you finally bring it home and plug it in, the network of motion sensors and light meters in the house immediately senses the torchire and turns it on-but only if it’s dark and you’re in the room. And if you decide to swap it with the lamp in the bedroom, no problem: the network figures that out as soon as you’re done.

In theory, linking together sensors, appliances and other devices so they can communicate and work together could make life easier and more productive. The reality-at least for now-is that setting up such networks is expensive and far from easy, especially if they involve thousands or even millions of components.

Now networks of devices that organize themselves-connecting to one another wirelessly and automatically, without human intervention-are moving out of research labs and into the marketplace. In their first incarnation, they will connect large numbers of sensors in factories and industrial settings, but within a few years they will move into office buildings, homes, even farm fields. Companies like MIT Media Laboratory spinoff Ember, Motorola and San Diego-based Sensoria are moving to create and sell the wireless radios and microchips that will enable devices ranging from temperature sensors to sprinklers to be connected in self-organizing networks [Technology Review board member Robert Metcalfe is an Ember investor and board member. Ed.]. “It really is the only form of networking that can work for lots and lots of little objects,” says MIT Media Lab researcher Michael Hawley. “The consequences of it really are going to be as magical as anything we’ve seen in technology.”

In a self-organizing network, says Ember cofounder and chief technology officer Rob Poor, “You just plunk these nodes down, and they discover each other and figure out how to get data back to where you want to get it to.” In other words, every element automatically recognizes every other element. Without any outside help, the devices must then determine how to get data where it needs to go.

One self-establishing network that is slowly making commercial headway is the Bluetooth wireless system. Originally designed by cell phone maker Ericsson to replace cables running between devices like computers and printers or cell phones and headsets, the system allows up to eight devices to connect to each other. When any device equipped with a low-cost Bluetooth radio comes within about 10 meters of any other Bluetooth-enabled product, the two automatically initiate a connection.

“Bluetooth is pretty well designed for what it was supposed to do, which was allow all the devices you carry around to talk to each other,” says Ember cofounder and chief scientist Andy Wheeler. But the eight-device limit and the design of the Bluetooth network both hinder the system’s usefulness for applications that require hundreds or thousands of devices spread over a large area.

Most wireless networks, including Bluetooth and the popular 802.11b “Wi-Fi” networks used to connect computers to the Internet, employ a spoke-and-hub organization in which one device acts as a central access point that all the other network members must communicate with directly. This turns out to be impractical in settings like factories, which are filled with machines, thick concrete walls and other radio-hostile things. That’s where Ember and similar network designs come in. With Ember’s networking protocols, the radio networks look like a mesh: every device recognizes other nearby devices immediately and can talk to all of its neighbors, passing data along. “Every node is a little bit of a router,” says Poor. “It sends the message on in the right direction.”

The initial application of Ember networks will be the replacement of expensive wiring in areas like factory floors. Other uses for the system will emerge as prices drop. For example, temperature, light and motion sensors could be placed in office buildings and networked to the lighting and ventilation systems. The network would then be able to tell when people were working in different areas of the building and switch on lights, heating or cooling only when needed. The same idea will eventually be applied to home automation, says Ember’s acting CEO Adrian Tuck. Tuck cites security systems as another emerging application. For example, networked biological-weapons sensors placed in air-conditioning ducts throughout a building or in water treatment plants could offer early warning of a terrorist attack (see “Networking the Infrastructure,” TR December 2001).

Motorola researchers see agricultural applications as well. Moisture sensors distributed throughout a field could be networked to irrigation systems, signaling the giant sprinklers to activate only when a part of the field was dry, instead of at regular intervals, saving water and money. The same scheme might be employed in a backyard sprinkler system.

Self-organizing architectures are also appearing for more complex networks. Engineers at IBM’s Almaden Research Center in San Jose, CA, are prototyping a data storage system made up of “collective intelligent bricks”: densely packed devices each consisting of a microchip, some memory and several hard-disk drives. Several hundreds of the bricks would be combined to create a single massive storage system. Software allows the bricks to recognize the addition of new bricks and figure out the best way to send data between them for storage. Similarly, if a brick fails, the system finds a way to route around it.

The goal of the brick system is to make storage servers simpler and cheaper to manage. Moidin Mohiuddin, the lab’s senior manager of advanced storage systems, estimates that one administrator can currently manage about one terabyte, or one trillion bytes, of data. He hopes a system composed of the bricks could increase that figure a thousandfold. Moidin says the architecture could work as well for other types of servers and ultimately even PCs, making setting up a home or office network as simple as turning on the machines.

As they make their way into more and more systems, self-organizing networks will do no less than transform the way we relate to everything from our computers to our appliances, making them, if not smarter, at least more helpful. “I think [the networks] will turn up in all sorts of creative ways,” says MIT’s Hawley. “The result is going to be a radical simplification of the way we interact with the stuff around us.”

Some Companies in Self-Organizing Networks

Company Application Sensoria (San Diego, CA) Home and office building automation; automotive applications Motorola (Phoenix, AZ) Wire replacement for factories; home and office building automation Palo Alto Research Center (Palo Alto, CA) Military applications and homeland defense MeshNetworks (Maitland, FL) Cellular replacement for mobile broadband voice and data services