AGENT
Overview
Project AGENT is exploring the broad range of non-physics-based modeling techniques to tackle key challenges in study for complex adaptive systems. While Agent-Based Modeling and Simulation (ABM/S) and System Dynamics (SD) remain at the core, other modeling techniques that are currently being leveraged and expanded as part of this effort include network models , discrete event simulations, constructive simulations as well as Markov models and stochastic Petri net models to name a few.
There is a growing awareness on the importance of complex adaptive systems (CAS) along with advances in modern sciences and information technology. The design of such a system-of-systems (SoS) is associated with a large number of entities and processes. To better understand and capture the complexity of these entities and their interactions (emergent behavior, non-linearity), a variety of methods and enabling tools should be employed depending on the nature of the problem. Further, a synergism of such methods and tools should be pursued to improve the efficacy of the solution approach.
For more information about INIT:AGENT, please contact John Lewe (j.lewe@gatech.edu)
Research Areas
Agent-based Modeling (ABM) and System Dynamics (SD) are widely accepted as representative approaches that can offer truthful perspectives on understanding SoS. In some cases SD is preferable, and in others ABM makes the most sense but the both approaches can also offer two different yet legitimate ways to look at the same problem. INIT:AGENT is hoping to not only understand and characterize these differences, but also looking for new, unique opportunities to investigate the combination of the two techniques in unison. The immediate goal is to achieve theoretical as well as empirical advances in this challenge.
Agent Based Modeling
Agent-based Modeling (ABM) is an approach to modeling systems and systems-of-systems as autonomous, interactive agents. There are differing opinions about what constitutes an agent; some require that an agent have adaptive behavior, but others believe that any independent component (e.g. surrogate model, person, etc.) is considered an agent. Agent-based Modeling is often referred to by several common names, including: Agent-based modeling and simulation (ABMS), Agent-based Simulation (ABS), and Individual-based modeling (IBM). Regardless of the difference of opinion on definition of an agent and the name of the process, ABM has several core principles that are widely accepted. One of the founding principles of ABM is that agents have the capability to make decisions. This can be achieved either through basic conditional rules (e.g. if-then statements) or through highly adaptive techniques (e.g. Artificial Intelligence). One of the main characteristics of ABM is that the technique is bottom-up by design. One of the main benefits of ABM is the ability to see complex behavior emerge from simple rules (i.e. emergent behavior). This can be seen from several well-known examples like the Game of Life and the Boids Model. In both games, three basic rules lead to complex behavior. In the Game of Life, complex patterns can emerge from random placement of agents, and these patterns can be sustained indefinitely. In the Boids Model, flocking behavior can be seen amongst the randomly dispersed agents.
System Dynamics
System Dynamics is another methodology used in System Thinking. It helps to understand the behavior of complex system over time. It is widely used to understand industrial processes, and for policy analysis and design. Due to the complexity of some system of systems, the identification of the components and the relationships between them is necessary. It deals with interdependence, mutual interaction, information feedback and circular causality. The System Dynamics Paradigm is based on the following principles:
- Linear analysis is not suitable
- Systems are unstable
- Predicting functions is possible
- Analysis should not be limited to existing data
- Model must not be limited to accepted definitions
- Descriptive knowledge is suitable for modeling
- Accuracy is not numerical accuracy
Every model starts with the identification of the main variables in the system and their interactions. Two representations are used to model the system: causal loop diagrams and stock and flow diagrams. The causal loop diagram (including feedback loops) defines variables and how they connect with each other. Stock and flows are used to represent material or immaterial accumulations and rates. System Dynamics is a top-down approach, as opposed to the bottom-up approach of Agent Based Modeling.
