Populating Simulation Environments

From Master Projects
Jump to: navigation, search

has title::Populating Simulation Environments
status: finished
Master: project within::Technical Artificial Intelligence
Student name: student name::Jan-Willem Boon
Start start date:=2012/01/01
End end date:=2012/08/01
Supervisor: Natalie van der Wal
Second reader: has second reader::Philip Kerbusch
Company: has company::TNO
Thesis: has thesis::Media:JCBoon-Thesis.pdf
Poster: has poster::Media:JCBoon-PopulatingSimulationEnviroments.pdf

Signature supervisor



The aim of the project is to create agents which act as people in a demonstration in an urban area. These agents have to react to peacekeeping forces which try to defuse the situation and stop the demonstration. Several options are available to the peacekeepers which range from telling them to stop to shooting them. The demonstrators will have to act naturally to these actions.

For the project we will decide on two different scenarios and see if we can model agents in such a way that they act naturally. To see if they act naturally we can find videos of demonstrations and the reactions of people and see if our agents would react in the same way as in the video.

In the end the agents will be used to train peacekeepers in finding the best way to solve demonstrations without losing lives and decreasing the morale of the civilians.

Abstract KIM 1

During the graduation periode the aim is to answer three questions:

  • How can a crowd of demonstrators be modelled using decentralized techniques?
  • Is a decentralized crowd more realistic then a centralized crowd?
  • How does the centralized crowd perform in Virtual Battlespace 2?

To answer these questions several steps have to be completed. These steps are the following:

  1. Research how a crowd of demonstrators work. Determining the interaction between the demonstrators, control forces and their non lethal weapons. This will be researched by doing a literature study.
  2. What decentralized methods exist to model the demonstrators. Research will be done in methods like BDI, Behavior Trees and Cellular Automata. This will also be done by a literature study.
  3. Determining if a decentralized crowd is better then a centralized crowd. To determine this an experiment will be set up in a 2x2 design and the results will be analyzed by using ANOVA.
  4. How does the decentralized crowd perform in VBS2? This will be done by measuring several aspects while running the decentralized crowd in VBS2.