Computing has evolved from highly centralized systems, organized around mainframes, to local area and wide area networks that tie together hundreds and even thousands of computers to the Internet, which is arguably the biggest network of them all. The next step, according to computer scientist David Evans, is swarm computing, involving hundreds of thousands of mote-sized devices spread throughout the environment.

As the name suggests, a close analog to swarm computing is an ant colony or beehive. All the members of these groups are independent, yet they organize and reorganize, as either individuals or groups, to take on different tasks required for the common good as the circumstance arises.

In a computer swarm, each tiny device might contain a processor, memory, transmitter, power source, and sensor. Spread outside a secure facility, the swarm might be used to identify intruders and to invoke appropriate levels of security based on the perceived threat. Creating such a massive network in which each of the components retains the potential for individual action while serving collective ends is a complex challenge.

Interested in finding a way to develop applications that would run on swarm computers, Evans has taken biology as his model. "The genome is a small program in size, but much more complex in effect than programs we currently produce today," he observes. "Incorporating principles from biology may help us develop more effective swarm programming."
For instance, in a developing embryo, independent cells react to changes in their environment and to chemical signals to change form, thus producing tissues and organs. Taking his cue from nature, Evans also focuses on the interactions among simple components, rather than more complex systems, as they self-organize to produce a desired objective under specific conditions.

Evans is further inspired by the robustness of most biological organisms, even in the face of a hostile environment that can destroy relatively weak individual components. Because individual elements in a swarm computer will be cheap and disposable, swarm computing networks must be self-healing and adaptable.

Evans is in the formative stages of this work. With support from an NSF CAREER award and an Information Technology Research (ITR) award, Evans and his research group have created a program to simulate cells' response to changes in their environment and to the failure of nearby cells. He is also building a library of simple actions, which are called primitives, that these individual elements might perform and developing ways to measure the robustness of the systems he creates. "My goal is to give programmers the tools needed to produce sophisticated behavior in swarms without having to worry about each individual device," he says.

This story first appeared in the Explorations, a publication produced by the Office of the Vice President for Research and Graduate Studies. Visit Explorations here.