Simulation and perception of variable density crowds

Abstract

Crowd simulation in virtual environment is of great utility in a variety of domains, from the entertainment industry to serious applications including social robots. However, current research in this field either focuses on single virtual characters, or macroscopic crowds. The link (small groups) between large crowds and individual virtual characters remains underexplored. In this seminar, my ongoing PhD research is presented in three interrelated and progressive parts:

1. experiments in mixed reality environments with individual humans and agents in order to determine initial parameters (concerning, for example, social distances) and nonverbal behaviors, as a basis for modelling efforts,
2. simulation of social behaviors of variable density crowds via machine learning methods, from collision avoidance and navigation in large crowds to group joining behaviors, and
3. perceptual studies are performed to investigate interactions at larger scales involving large crowds.

Moving forward, we aim to extend small group behaviors to a larger crowd. This research finds its importance in virtual agent scenarios and real situations with physical robots which usually involve interactions across the individual, group and crowd levels.