Velocity-Space Reasoning for Interactive Simulation of Dynamic Crowd Behaviors Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 19, 2019
Creator
  • Kim, Sujeong
    • Affiliation: College of Arts and Sciences, Department of Computer Science
Abstract
  • The problem of simulating a large number of independent entities, interacting with each other and moving through a shared space, has received considerable attention in computer graphics, biomechanics, psychology, robotics, architectural design, and pedestrian dynamics. One of the major challenges is to simulate the dynamic nature, variety, and subtle aspects of real-world crowd motions. Furthermore, many applications require the capabilities to simulate these movements and behaviors at interactive rates. In this thesis, we present interactive methods for computing trajectory-level behaviors that capture various aspects of human crowds. At a microscopic level, we address the problem of modeling the local interactions. First, we simulate dynamic patterns of crowd behaviors using Attribution theory and General Adaptation Syndrome theory from psychology. Our model accounts for permanent, stable disposition and the dynamic nature of human behaviors that change in response to the situation. Second, we model physics-based interactions in dense crowds by combining velocity-based collision avoidance algorithms with external forces. Our approach is capable of modeling both physical forces and interactions between agents and obstacles, while also allowing the agents to anticipate and avoid upcoming collisions during local navigation. We also address the problem at macroscopic level by modeling high-level aspects of human crowd behaviors. We present an automated scheme for learning and predicting individual behaviors from real-world crowd trajectories. Our approach is based on Bayesian learning algorithms combined with a velocity-based local collision avoidance model. We further extend our method to learn time-varying trajectory behavior patterns from pedestrian trajectories. These behavior patterns can be combined with local navigation algorithms to generate crowd behaviors that are similar to those observed in real-world videos. We highlight their performance for pedestrian navigation, architectural design and generating dynamic behaviors for virtual environments.
Date of publication
Subject
DOI
Identifier
Resource type
Rights statement
  • In Copyright
Advisor
  • Manocha, Dinesh
  • Guy, Stephen
  • O'O'Sullivan, Carol
  • Frahm, Jan-Michael
  • Lin, Ming
Degree
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2015
Language
Publisher
Place of publication
  • Chapel Hill, NC
Access
  • There are no restrictions to this item.
Parents:

This work has no parents.

Items