From the earliest days of computers, engineers have been fascinated with how computation happens in natural systems, and how digital circuits can mimic biological mechanisms. At first, the focus was on what happens inside a single brain. Both connectionist and symbolic approaches to machine intelligence appealed to biological realism as justification for their approaches. Since the mid-1980s, researchers have paid increasing attention to computation that takes place across a population of autonomous but interacting processes. When this research draws explicitly on a biological model, it is commonly known as swarm intelligence.
National Geographic popularized research in this area in July 2007, with the feature article, “The Genius of Swarms,” by senior editor Peter Miller. Now, Miller extends that exposition into a book-length popular treatment that outlines several different forms of swarm intelligence, summarizes the biological models on which they are based and the research that has elucidated them, and discusses how these patterns of animal behavior contribute to the human decision-making process, both with and without the help of computers.
Miller organizes his discussion around five animal models: ants, honeybees, termites, birds, and locusts. The ability of ants to plan paths by depositing and sensing chemicals in their environment is one of the best-known mechanisms in swarm intelligence. Positive feedback among successful foragers, combined with evaporation and propagation in the physical environment, leads to the emergence of large-scale structures from many individual contributions. These mechanisms have been extensively exploited computationally to solve industrial problems of routing and scheduling.
Honeybees face a complex problem in selecting a nesting site. This single decision can make the difference between life and death for the hive, yet individual bees are too small to review multiple alternatives. For over 60 years, biologists have known that bees use a highly structured “waggle dance” to communicate information about rich sources of nectar and pollen. Recently, biologists have found similar mechanisms that encode reports from individual scouts about promising nesting sites. The integration of these diverse reports bears a striking resemblance to mechanisms by which people integrate information from multiple experts, or even multiple ordinary people with different perspectives, as in a town meeting.
Termite mounds are impressive engineering feats, with multiple floors and chambers, and complex air-conditioning duct systems that use solar-powered convection to remove gaseous wastes and bring in fresh air. The construction of such structures draws on chemical signals that are related to ant pheromones, but that guide a much more complex task than planning point-to-point paths over a two-dimensional surface. This process is analogous to how people interact in a Web 2.0 information environment such as a wiki. In social settings, the network of connections among people is an important part of the framework for such information integration. In engineering settings other than networks, such as a power grid, it can serve a similar function.
The coordinated movements of large flocks of flying birds have attracted attention for decades, but only in the past few years have advances in computer vision made it possible to derive the rules of motion of individual birds from which these large patterns emerge. This coordination serves the birds in communicating information about nearby predators that are faster than a single bird can fly. It also makes it difficult for a predator to single out a victim. Imitation of these algorithms has led to revolutionary advances in animation for motion pictures, and holds promise for the efficient control of teams of robots. It also helps explain herding behavior among people.
Complex group behavior that emerges from simple individual actions holds considerable engineering promise, but can also be counterproductive. Miller presents locust swarms as prototypes of swarming gone wrong, and compares them to stadium riots and financial bubbles.
This volume is an informal, highly accessible summary of how different forms of coordination among multiple animals can inspire both computational and social actions by humans. Interested readers will want to follow up with the references provided, in order to learn more about the individual research projects it mentions, or perhaps with more technical volumes [1,2].
By providing a high-level overview, this volume captures the exciting multidisciplinary nature of the work in the field better than any single specialized study, as research by biologists, psychologists, and computer scientists cross-fertilizes at multiple levels. In addition to introducing these ideas to the general public, the volume could prove very valuable in a high school curriculum, as a way to expose students to the interplay between biological and computational research, and, perhaps, stimulate the next generation of researchers who will extend these mechanisms further.
