Back to search

University of Sheffield

Constraint-based simulation of virtual crowds

Abstract

dc:description.abstract

Central to simulating pedestrian crowds is their motion and behaviour. It is required to understand how pedestrians move to simulate and predict scenarios with crowds of people. Pedestrian behaviours enhance the range of motions people can demonstrate, resulting in greater variety, believability, and accuracy. Models with complex computations and motion have difficulty in being extended with additional behaviours. This is because the structure of these models are not designed in a way that is generally compatible with collision avoidance behaviours. To address this issue, this thesis will research a possible pedestrian model that can simulate collision response with a wide range of additional behaviours. The model will do so by using constraints, a limit on the velocity of a person's movement. The proposed model will use constraints as the core computation. By describing behaviours in terms of constraints, these behaviours can be combined with the proposed model. Pedestrian simulations strike a balance between model complexity and runtime speed. Some models focus entirely on the complexity and accuracy of people, while other models focus on creating believable yet lightweight and performant simulations. Believable crowds look realistic to human observation, but do not match up to numerical analysis under scrutiny. The larger the population, and the more complex the motion of people, the slower the simulation will run. One route for improving performance of software is by using Graphical Processing Units (GPUs). GPUs are devices with theoretical performance that far outperforms equivalent multi-core CPUs. Research literature tends to focus on either the accuracy, or the performance optimisations of pedestrian crowd simulations. This suggests that there is opportunity to create more accurate models that run relatively quickly. Real time is a useful measure of model runtime. A simulation that runs in real time can be interactive and respond live to user input. By increasing the performance of the model, larger and more complex models can be simulated. This in turn increases the range of applications the model can represent. This thesis will develop a performant pedestrian simulation that runs in real time. It will explore how suitable the model is for GPU acceleration, and what performance gains can be obtained by implementing the model on the GPU.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Sheffield
Year dc:date.issued
2023

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Charlton, John Anthony
Advisors dc:contributor.advisor
  • Richmond, Paul
  • Maddock, Steve
ORCID person:identifier.orcid

Rights

Language dc:language.dcterms_iso639_2
eng

Identifiers

dc:identifier.*
Dc Identifier Grantnumber
EP/N018869/1

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Charlton, John Anthony. Constraint-based simulation of virtual crowds. doctoral thesis, University of Sheffield, 2023.