Agent-based models

The fundamental building block of most models of complex adaptive systems 

In general, an adaptive autonomous agent is cliaracterized by the following properties [maespQO]  

Since a major component of an agent s environment consists of other agents, agents generally spend a great deal of their time adapting to the adaptation patterns 

Insofar as complex adaptive systems can be regarded as being essentially open-ended problem-solvers, their lifeblood consists mostly of novelty. The ability of a complex adaptive system to survive and evolve in a constantly changing environment is determiimd by its ability to continually find — either by chance, or experience, or more typically both insightful new strategies to increase its overall fitness (which is, of course, a constantly changing function in time). 

---

Any system whose top-level behavior is a consequence of the aggregate behavior of lower-level entities - biological systems, neural systems, social systems, economic systems, among many others - can, in principle, be simulated by an agent-based model. 

Epstein [epsj99] presents a cogent argument for why agent-based models are particularly useful as generative, explanatory, tools in social science. 

In this section we introduce a mobile CA model of land combat called EINSTein, developed at the Center for Naviil Analyses for the US Marine Corps. We include a discussion of this model here becau.se it is an interesting blend of CA-like local dynamics and agent-based modeling techniques. 

This example illustrates that when the resources and personalities of both sides remain fixed in a conflict, how well side X does over side Y does not necessarily scale monotonically with X s sensor capability. As one side is forced to assimilate more and more information (with increasing sensor range), there will inevitably come a point where the available resources will be spread too thin and the overall fighting ability will therefore be curtailed. Agent-based models such as EINSTein are well suited for providing insights into more operationally significant questions such as, How must X s resources and/or tactics (i.e., personality) be altered in order to ensure at least the same level of mission performance ... 

Payoffs to Using Agent-Based Models of Combat... 

The payoff of an agent-based approach is a radically new way of looking at fundamental issues of land combat. Models such as these are emphatically not to be used for prediction rather they are best used to enhance understanding. Specifically, agent-based models of combat are designed to help military theorists... 

Any colony optimization (ACO) and swarm intelligence are forms of agent-based modeling inspired by colonies of social animals such as ants and bees [32]. ACO has become popular in engineering for optimal routing in water distribution systems [33, 34]. Particle swarm optimization has been successfully used to train ANNs, for instance, ANNs to predict river water levels [35], for parameter estimation, for example, in hydrology [36]. 

Several other types of evolutionary algorithms (EAs) exist, differing in the selection method for parents and the way in which one population is formed from the previous one or in the interpretation of the GA strings. In addition, evolutionary agent-based models are starting to appear. Although these methods show promise, they are currently less widely used in science than the GA, so are covered here briefly. 

Agent-based models create individual enhties in the dynamic modeling framework to represent unique chemical or biologic compounds. 

In a variety of ways, the sciences of complexity are coming to bear on economics. For example, an agent-based model of a stock market has been constructed and it has shown that both fundamental pricing behavior and speculative bubbles can occur. 

Topping CJ, Hansen TS, Jensen TS, Jepsen JU, Nikolajsen F, Odderskaer P. 2003. ALMaSS, an agent-based model for animals in temperate European landscapes. Ecol Model 167 65-82. 

J. M. Epstein. 1999. Agent based models and generative social science. Complexity 4(5) 41-60. 

For pesticide risk assessment, 3 major types of population models can be distinguished (Bartell et al. 2003) difference or differential equations, matrix models, and individual- or agent-based models. Within each type, further distinctions can be made, for example, regarding the inclusion of stochasticity or spatial effects. However, these distinctions are less fundamental than the choice of the model type itself. 

Pertoldi C, Topping C. 2006. Impact assessment predicted by means of genetic agent-based modeling. Crit Rev Toxicol 34 487M98. 

---