چكيده به لاتين
Crowd simulation is the process of simulating the movement of a large number of entities or characters. It's most important applications are in fields such as entertainment (computer games), emergency evacuation (for building safety testing) and traffic control (crowded places like the metro station). This type of simulation is related to many issues. The most important of these are computer graphics, the population behavior modeling, collision avoidance and routing. In this thesis, our focus is on behavior modeling.
For the first time in 2005, a method was proposed to support real world data to rebuild a crowd behavior. However, for some reasons, these methods were not used in practice. To be more elaborate, low precision of the methods of tracking individuals and low computing resources for real-time simulation were two main drawbacks. However, in recent years, due to the advancement of technology and the availability of suitable data for this, data-driven methods are most popular among the methods of crowd simulation.
In this thesis, we have presented a data-driven method for crowd simulation, in which, with the help of the concept of holon, we could increase the accuracy of simulation and bring its output closer to a real crowd. To prove this, we first prove that crowd is an example of a holonic structure, and this point of view increases the precision of work. Then, using real-world data, the rules of joining each agent into a holon and separating from it are modeled using a classifier, which helps us model it in simulation. Also, because we used data from a specific environment, we tested the model with data from another environment, and it was found that the rules derived from the first environment were largely existed in the second environment. This result confirms the generality and comprehensiveness of the proposed method.
Keywords: crowd simulation, data-driven model, holonic multi-agent systems
