Introduction

In this article, general concepts such as model and multi-agent system are described. A general understanding of these concepts is essential to being able to design effective models and work with the MARS systems.

What is multi-agent modelling?

A model is a simplified representation of some specific aspects of reality. The goal is creating a model is typically to define and understand a concrete information system that closely resembles its real-world counterpart. When working with and running a model (in case of MARS, in the form of simulations), the obtained output data can be used to inform decisions in the real world.

Agent-based modelling derives from the field of artificial intelligence (AI). This simulation paradigm incorporates individuals, so-called agents, who interact with each other and their surroundings. The behavior is programmed on an individual level to follow a set of rules: The interactions between agents that occur as a result of individual behavior are studied to gain insights into collective behavior. Note that an agent is not restricted to be an individual but can also be a group, community, or other entity that acts and reacts to external conditions.

The way of creating results bottom-up from an individual's actions leading to complex effect makes multi-agent modelling especially well-suited for research in the social sciences.

Tick-based simulation are in contrast to continuous simulations divided into steps. These steps symbolize a time-driven progress. Every tick stands for an equally large time frame, which is called $\Delta t$. This time frame is fixed for the whole simulation, so that for example a simulation is progressing in steps of 1 minute. Every scenario runs for a defined amount of ticks. In every tick all tick-based elements (agents and active layers) have the possibility to act. Time-referenced scenarios have a start and an end time-point and a step length ($\Delta t$). The amount of ticks is then inferred.

What is MARS?

MARS is a software framework written in object-oriented C# that enables the development and simulation of ABMs.

The MARS runtime system provides agent-based simulation and geographical query processing methods with a set of standard mathematical functions and data structures for .NET Core.

The framework provides simulation core to execute the simulation with variable step-size. It is compatible with multiple platforms including Microsoft Windows, macOS, Xamarin, Unity3D, Windows Store applications, Linux or mobile.

The framework offers a unified API to develop and execute individual-based models and scenarios.

References

See the following video, for what's possible with MARS:

Also see some more complex models regarding mobility in urban settings.

For a more simple model have a look at our Wold-Sheep-Model for an introduction into what's possible with ABM, to simulation a predator-prey dynamic.

Also have a look at further models at our GitHub account.