Psychology-based Crowd Simulation
IntroductionThis project proposes a model for simulating human behavior in crowds, heavily based on psychology.
The model accounts for individual agents with their own, independently perceived environment and responses to said perceptions. At its core, it's an AI capable of dynamic collision avoidance on a basis of proxemics, other perceived entities and static obstacles. The model is also capable of navigating unknown environments by making use of target flow.
For the realism aspect, the AI is influenced by psychological behavior as well as have an identity model to mimic personality.
ResearchFor a period of 4 weeks I studied theoretical psychology to learn all about stress and its mathematical applications, proxemics and developments like the general adaption syndrome.
I also studied the influence of identity and personality on behavior, and sociological occurrences like authority and deindividuation.
Furthermore, I studied Helbing's social force model and searched for ways to improve on it.
DevelopmentThe crowd simulation is agent-based, so I built one agent and proceeded to use that agent to construct a crowd.
The agent is capable of dynamic collision avoidance and wants to maximize its degree of comfort at all times while also accomplishing its goal.
The behavior of the agent is fundamentally defined by the social force model, which, in this implementation, consists of driving forces, social forces and repulsive stress forces.
In addition to calculating and being influenced by positional stress and interpersonal stress, the AI also applies a form of herd behavior.
The identity model is one I composed myself, which I called the employed artificial identity, and it guarantees that every agent behaves in a unique fashion. It consists of 5 key attributes: dominance, agreeableness, anxiety, mobility and authority.