Wednesday, September 5, 2012

Think Like an Ant

The title of this blog post is slightly misleading, because I'm finding that to "think like an ant" would mean almost not at all. Many species of insects, such as ants, bees, and termites, form social colonies often characterized by one reproductive individual (the queen) and hundreds to millions of non-reproductive workers (in bees and ants, all these workers are female). Because these colonies in some ways essentially form a single distributed individual, they are often referred to as "SuperOrganisms".

These are complex societies, complete with clear divisions of labor, the ability to exploit new food sources and defend the nest from enemies, and the capacity to move the entire community to a new nest location if necessary. But there is no centralized "command and control" by the queen or any other individual.

In place of any centralized control are a set of highly-evolved behaviors that each individual is essentially born with. Ant workers begin these tasks from the moment they emerge as recognizable adults, often immediately beginning to care for nearby young in the brood chamber. No on-the-job training, no instruction from mission control, no learning. They simply operate.

Often worker activities are guided by a simple algorithm selecting from a small set of choices based on situation. For example, an ant worker encountering a hungry larva in the brood chamber feeds it. If the same ant encounters the same larva outside the brood chamber, she carries it back to the brood chamber, hungry or not. A relatively small set of these highly-evolved simple decision trees is multiplied by thousands (or even millions) of workers and then integrated in the form of the colony to result in a highly adaptive, seemingly very intelligent SuperOrganism. But the actual decision-making process of any individual worker is very simple.

These basic principles of individual simplicity and collective sophistication could be mirrored in the structure of the Internet of Things. End devices may be outfitted with only the barest of intelligence, simply broadcasting their state or responding to simple direction from other functions. Meanwhile, intermediate devices in the Internet of Things may be forwarding messages as necessary to effect communications throughout the system.

A key to keeping the overall architecture simple may be gleaned from the uncomplicated social insect decision process outlined above. A basic set of communication and networking rules, multiplied billions of times over in devices throughout the Internet of Things, may yield great sophistication in overall function. Each individual ant does very little "computing" – the same may be true for many members of the Internet of Things.

One other factor has led to the incredible success of many social insect colonies: the relatedness of the workers in the colony. In many cases, the workers are all sisters of the same parents. This makes them very closely related to one another. Because of this (and because workers typically don't reproduce), there is no incentive for competition with nest mates. Thus, the various classes of workers (or castes) perform a variety of duties on one another's behalf selflessly. No external control is needed to guide the colony to maximize its environmental fitness.

In contrast, the Internet of Things is being architected by Homo sapiens, whose innate drive for competition with one another is adaptive and tragically well-documented. If this competition (one might label it "greed" if uncharitable) is reflected in the architecture of the Internet of Things, a great opportunity will be lost.

If instead devices in the Internet of Things can be designed to "help out" other devices (when the costs to do so are low); there will a tremendous benefit to the network as a whole. This might take the form of forwarding messages for unrelated devices or other means of insuring better overall function at a minor cost of delay or compute power. Thinking like the humble ant might prove to have great power in the Internet of Things.

Some writers have used ideas from nature simply as metaphors for the IoT, while still others make a more meaningful link to network engineering. I believe that the latter are on the right track: architecting the Internet of Things based on successful natural models of massive systems offers many possibilities.

In the next posts, I may move off communications for a while and explore other areas of the IoT.