[ Reply | Next | Previous | Up ]
From: Josephine Maisonet
Date: 4/30/01
Time: 9:32:23 PM
Remote Name: 24.168.112.94
SWARM SMARTS
Modeling the social organizations and behavior of ants and other social insects, computer scientists have been able to create software agents that mimic these social insects. An agent is something that perceives and acts. In the field of artificial intelligence, one of it's aims is studying and construction of rational agents such as robots that are programmed to do certain tasks.
But why study and model after such insects that live in colonies? They are fascinating in their own right, but how can this knowledge be any use to computer scientists? Why do the following questions need to be answered in respect to insect colonies?
What is it that issues orders? How does it foresee the future? How does it elaborate plans and preserves equilibrium?
Apparently a humongus amount. These questions do not have immediate and available answers; indeed, these questions are quite puzzling to scientists. By observations, it has been shown that each insect colony seems to have its own agenda, yet the whole group appears highly organized. Each insect's individual activity does not require any supervision yet the integration of all the individual activities is seamless and efficient.
Scientists who studies these types of insect behavior have found that cooperation at the colony level is largely self-organized. In numerous situations, the interactions among individuals gives rise to the coordination of activities.
For example with ants a simple interaction amongst individuals in the colony (leaving a cent trail left by another) working together they solve the problem of finding the shortest route to a food source among the countless possible paths that can be taken.
This type of collective behaviorism has been called "swarm intelligence". There is a growing community of researchers that have been devising innovative ways of applying swarm intelligent to many and diverse tasks and problems. They have applied foraging of ants' methodology to the rerouting of network traffic in busy telecommunications systems. The cooperative interaction of ants has led to the development of more effective algorithms for robots. Mimicking a colony's way in dealing with its dead and sorting of their larvae has aided in analyzing banking data. In streaming assembly lines in factories, help has come from the division of labor among honeybees.
In one of the earliest studies of swarm intelligence, it has been shown that ant "highways" often seen in nature (even in your kitchen) result from having individual ants exuding pheromone, a chemical substance, that attracts the other ants. One of the researchers in particular, Jean-Louis Deneubourg, demonstrated that the process of laying pheromone trails led to a good strategy in finding the shortest path between a nest and a food source. In a computer simulation of pheromones, and pheromone evaporation computer scientists have developed software ants mimicking pheromone trails and pheromone decay. These software ants have been able to select a shorter branch. This property is highly desirable in that it prevents the system from converging on mediocre solutions.
Other swarm intelligence techniques have been applied to a number of optimizing tasks successfully. These software agents cooperate together to solve complex problems. One such complex problem, which has been successfully solved by insect software agents, are the re-routings of traffic in busy telecom networks. The artificial ants have also provided the best solution to the classic quadratic assignment problem, in which the manufacture of a number of goods must be assigned to different factories so as to maximize the total distance over which the items need to be transported between facilities.
Ant teamwork have inspired scientist to program robots without the use of complex software. In the University of Alberta, the researchers created robots that push an illuminated circular box toward a light. Each robot does not communicate with others, and they act independently by just following a small set of simple instructions. yet together as a group they're able to accomplish its' goal. The extreme simplicity of this ant-based approach (the robot does not need to communicate with each other) promises for miniaturization and low-cost applications.
The potential of swarm intelligence is tremendous. It offers alternative ways of designing systems that have traditionally required centralized control and extensive preprogramming. Swarm method " boats autonomy and self-sufficiency", relying on direct and indirect interaction and cooperation among simple individual agents. These types of operations could lead to systems that can adapt and learn quickly to rapidly fluctuating and dynamic conditions.
But the field is in its infancy. Researchers still lack in-depth knowledge of the inner workings of insect swarms. Without knowing the rules by which individual insect act under, it is difficult for scientist to develop appropriate algorithms and software. As the computer scientist Kevin Kelly stated, "Dumb parts, properly connected into a swarm, yield smart results". The challenge of course, is in the proper connection of all the parts.
Work Cited:
Bonabeau, Eric Theraulza, Guy
"Swarm Smarts"
Scientific American March 2000, 73-79