Discussion papers

CPM-14-225 - 5 December 2014

Agents in Space and Time

Ruth Meyer

Meyer, Ruth (2014): Agenten in Raum und Zeit — Diskrete Simulation mit Multiagentensytemen und expliziter Raumrepräsentation. PhD Thesis, Hamburg University, Germany.
URN: urn:nbn:de:gbv:18-70440
URL: http://ediss.sub.uni-hamburg.de/volltexte/2014/7044/pdf/Dissertation.pdf

This PhD Thesis is written in German.

Abstract

The last ten years have spawned a plethora of so-called agent-based models. This type of model is characterized by mapping entities and their interactions in the system under observation directly onto entities in the model (“agents”) and interactions between agents, respectively. Thus the model preserves the structure of the modelled system while investigating its behaviour over time. The wide range of application domains includes sociology, economy, biology, ecology, archaeology and the military sector.

Since these domains often contrast the agent-based style of modelling with continuous simulation, in which a set of differential equations is solved numerically by stepwise integration, it is not surprising that the vast majority of agent-based models apply a time-driven approach, i.e. simulation time is advanced in equidistant steps. This time advance method is considerably easier to implement than the more flexible and efficient event- driven approach, which can make all the difference if a dedicated simulation toolkit is missing. Both applications and toolkits are usually developed by domain experts and not simulation experts.

The present dissertation thesis shows how agent-based simulation can be defined as a new world view within the context of discrete-event simulation. Named “agent-oriented” in accordance with the (German) terms for the classical world views, this world view combines the metaphor of multi-agent systems with an event-driven time advance. Applying the event-driven approach requires that (a) the durations for agent and environment actions are determined before they terminate so that the respective termination event can be scheduled in time, (b) each agent is able to instantly react to changes in its environment, and © the update of the state of the environment can be kept efficient despite updating agents asynchronously.

Based on these requirements the concept for a simulation toolkit was developed, which implements the agent-oriented world view. A major focus of this toolkit is to support an explicit representation of space within the environment and the movement of agents in that space. These are areas where existing toolkits for agent-based modelling show shortcomings, despite the fact that a majority of multi-agent models explicitly model space and allow for mobile agents.

A prototypical implementation of this concept is provided with the simulation framework FAMOS extending the tried and tested discrete-event simulator DESMO-J. FAMOS was first tested by re-implementing two well-known models (Schelling’s segregation model, Epstein and Axtell’s Sugarscape model) and then thoroughly evaluated in a practice-oriented simulation study of a city courier service.

Downloads