Home - csteven1/MultiAgentSimulator GitHub Wiki

Overview

This tool is an open-source Java based GUI software application for creating multi-agent systems. This movements between spaces undertaken in this system are Markovian, so historical movements have no influence on the current movement possibilities.

The specifications of the desired system is entered by the user and the application carries out the simulation, saving the results to a CSV file for the user to view.

Note: This application was originally designed for modelling human occupancy in buildings, hence agents are referred to as people and spaces referred to as rooms. This application still works just the same for any other multi-agent system purpose, provided the modelling basis is Markovian.

The options for the user to configure are:

  • The number of rooms in the system.
  • The number of people in the system.
  • The length of the dataset, i.e. the number of time increments.
  • The movement probabilities between rooms.

Using these specifications, the simulation moves each person to another room, and where they move is dependent on the movement probability to each of the other rooms given where they are currently located, each of which has been specified by the user. When this has been performed for each person in the system, we increment to the next time-period and the same is performed for each person's new location.

In the current version, every person in the system starts in room 1 at the first time-step.

This application has been used successfully in the development of an occupancy prediction model for refining movement probabilities over time.

For further information on running the application, follow the instructions in this Wiki.