Agent-based Crowd Simulation Modeling in a Gaming Environment

摘要

Human crowds are often studied by experts from various areas for planning, training, investigation, etc. Furthermore, simulating the motion of crowd for realistic animation is a popular subject in the computer graphics and video games communities. Thanks to improvements in computing power, modeling and simulations, using computers, are now used to investigate the dynamics of crowds using agents. This paper proposes a novel agent-based crowd simulation model, using a game engine. Compared to the existing techniques, our proposed model is more flexible and, has the ability to make real-time changes in crowd-modeling. The agent generation method, known as Distinguishable Agents Generating Method (DAGM), are capable of creating 3D animated-humanoid models with distinguished attributes. In this paper, we also propose a Multiple layer Collision System (MCS) that includes a system to collect collision messages and to evaluate the processing. A Building & City-planning Generating System (BCGS) is also introduced to create and setup obstacles in the simulation. This approach can be used for studies, planning and training related to crisis management such as stampede, evacuation, etc. The proposed method is capable of rendering a visually vivid 3D crowd of hundreds of agents using a typical computing environment.