1986

SWARM INTELLIGENCE

Termite mounds can reach over 17 feet (5m) in height, with termites acting like simple “novelty detectors,” responding to changes in air characteristics within the mound and altering the structure as needed. Authors Doris and David Jonas speculate: “What is the other way of knowing, by which the termites know what they have to do and when they have to do it? Instructions cannot be brought to them quickly enough by messengers, since the distances within the hill are far too great. . . . A group brain functions as an instrument for decision-making in a way startlingly like an intelligent individual brain.”

The apparent collective intelligence in these social insects, as well as that in flocking and herd animals, inspired the concept of swarm intelligence. In AI, this concept is employed to deal with a range of challenges. The software agents follow simple, local rules and, as with ants and termites, there is no central controller dictating the behavior of the collective. As just one example, Boids, an artificial life program developed in 1986 by computer scientist Craig Reynolds (b. 1953), simulates the flocking behavior of birds by following simple rules involving a bird steering toward the average heading of birds in the flock, steering to move toward the average position of birds, and steering to avoid excessive crowding.

One of the numerous swarm methods studied in AI today is ant-colony optimization, a method using simulated ants that records their positions and quality of solutions to help ants in the colony determine better solutions. In some implementations, these “ants” simulate attractive chemical trails (pheromones) that evaporate over time. Particle-swarm optimization simulates the position and velocity of a swarm of fish toward an optimal location. Other interesting methods are artificial immune systems, bee-colony optimization algorithms, firefly optimization algorithms, bat algorithms, cuckoo searches, and roach-infestation optimization.

The applications of swarm intelligence include control of autonomous vehicles, routing in telecommunication networks, scheduling of aircraft, art creation, enhancing systems for reactive power and voltage control, and clustering of gene-expression data.

SEE ALSO Machine Learning (1959), Living in a Simulation (1967), Genetic Algorithms (1975), Artificial Life (1986), “Elephants Don’t Play Chess” (1990)