FAQ: Agent-based modeling of complex, adaptive systems

What are "complex systems" in this context? These are noncomputer systems, such as a company's supply chain. A system is "complex" when it has so many variables and interacting forces that it can't be understood in its entirety or optimized by traditional, top-down approaches.

How can you tame this complexity? Although these systems are complex overall, they use a few simple rules at local levels. For example, in a supply chain system, a rule in a warehouse might be, "Fill orders on a first-in, first-out basis," or "Don't send this truck out on delivery until it is full." Dozens or hundreds of these local "agents" - truck dispatchers, say - acting autonomously produce complex behavior by the system as a whole. It's possible to simulate this complex behavior by programming software agents with a few rules and letting them interact with one another. By optimizing the agents' activities at a local level, it's possible to improve the performance of the system as a whole.

Why are these systems called "adaptive," and why are they sometimes likened to ant colonies? Ants individually have extremely primitive brains, yet collectively they run surprisingly sophisticated and efficient operations. With no central direction, they divide responsibilities among themselves, find food, build and maintain their nests, tend to their young and respond to attacks. And the colonies adapt; if you block access to a source of food, ants will find an alternate route to the food. Complex adaptive systems do the same. For example, if Plant A can't satisfy a customer order because it's temporarily out of a raw material, Plant B may fill the order. Plant B may do this "automatically," based on simple local rules without direction from a central authority.

What is meant by "emerging behavior"? Like ants, individual agents can modify their rules to adapt to changing circumstances, and this can alter the global behavior of the system, often in unpredictable ways. Sometimes small, local changes can have big system impacts, just as a tiny disturbance in the atmosphere over Africa can lead to a hurricane days later in the Gulf of Mexico. Agent-based modeling can help us understand and predict these emerging behaviors and help us devise new rules for the local agents that will improve the performance of the system as a whole.

Copyright © 2003 IDG Communications, Inc.

  
Shop Tech Products at Amazon