Computer generated worlds are often required to present large crowds of entities to reduce number of dummies during movie production, to populate worlds in computer games or to create and test evacuation plans. Depending on application context the behavioral models used in these simulations should be elaborated and internal and external constraints should be taken into account.
The key component of the simulation is the effective description of the autonomous behavioral model of the entities and their interactions with each other and the simulated world. The simulation of the crowd dynamics should be based on the entity level behavioral model to produce the most realistic results. This approach makes developing different behavioral patterns for entities (like leader and followers) possible.
Current graphics cards' performance lets parallel algorithms to be ran in real-time, so it can be used to solve the naturally parallel problem of entity-level simulation. The goal of this thesis work is to produce a software capable of real-time simulation and visualization of large crowds. The software will allow observation of the dynamics of the crowds and inspections of different behavioral patterns.