The interest in crowd simulation has increased over the last years, mainly because of the increasing computer processing power that began to allow simulations with thousands of agents with plausible behaviors in real time combined with the interest of many areas in providing realistic simulations (e.g., entertainment, building analysis, studies in emergency evacuation, collaborative virtual environments).

2011 Teofilo_Crowd_01

 Crowd simulation is a computationally expensive task, where the system has to manage a lot of resources to provide a believable simulation to the user. This leads the systems to have to deal with the well known trade-off between realism and performance. 

This area is multidisciplinary. In a multi-agent system, each individual has its own goal and intentions to follow, and must to steer to it while avoiding collisions with other individuals and the obstacles in the environment, at this moment, the system has to handle path planning and collision avoidance algorithms that may be influenced by psychological aspects. These agents may also present social behaviors. As increasing the number of agents, it is necessary to worry about the performance, and, to handle this, the user should use parallelism techniques to divide tasks among processors (CPU/GPU) and rendering techniques, such as level of detail and image-based rendering, to render thousands of agents.

2011 Teofilo_SVR

 As one might suspect, this area has many open problems and some of them are described in [1].

In our group, the following students are researching in this area:

 More information about crowd simulation can be found in:

[1] Daniel Thalmann, Helena Grillon, Jonathan Maim, and Barbara Yersin. Challenges in crowd simulation. In Proceedings of the 2009 International Conference on CyberWorlds, CW ’09, pages 1–12, Washington, DC, USA, 2009. IEEE Computer Society.

[2] Daniel Thalmann and Soraia Raupp Musse.  "Crowd Simulation". Springer London, 2008.