Department of Computer Science at UH

University of Houston

Department of Computer Science

In Partial Fulfillment of the Requirements for the Degree of
Master of Science

Xiao Cheng

Will defend his thesis


GPU-accelerated Crowd Simulation with User-Guidance and Multi-level Uncertainty

Abstract

A long-standing rule for evaluating the realistic and robustness of the crowd simulation is whether the simulations are capable of resolving the congestion as well as avoding agent-wise collision. Not a well-rounded approach has been exploited yet to deal with all real-world crowd scenarios.

This work is motivated to propose a new framework to simulate large-scale and realistic crowd behaviors with informal cultural convention. Technically, we construct a hybrid crowd advection pipeline by combining the global planner and local planner. To the best of our knowledge, a two-level uncertainty model is prestented, first-of-its kind, and imposed on the above pipeline to reflect microscopic variability upon macroscopic similarity. A user-interactive navigation scheme is incorporated to resolve pedestrian traffic congestion and form cultural convention. Lastly, a GPU-friendly local collision avoidance method is utilized to avoid collison artifacts among people.

All simulation and visualization are accelerated by harnessing GPU. Through our experiments and comparisons, we validated the advantages of our approach over a few other works.

 

Date: Monday, November 4, 2013
Time: 2:30 PM
Place: PGH 550

Faculty, students, and the general public are invited.
Advisor: Prof. Zhigang Deng