Fuzzy finite state machines in crowd simulation

Large crowds of pedestrians are a common phenomenon in big cities, this field of research studies and reproduces this kind of situations in virtual environments. Complex interaction within the agents is desired but it requires modeling their internal state. Through the internal state of the agents we can include reactive behaviors with multiple and changing objectives. The individual (personality) and social (group membership) properties of the agent can also be modified due to external or internal changes, such as environment, communication and mood. The inclusion of this features, and the internal state of the agents, change the behavior of the crowd. One of the main problems when simulating a crowd is to define individual traits for all the agents within the simulation. It is possible to define parameters that individually describe each of the characters involved in the simulation, however this approach turns to be unfeasible as the crowd grows bigger. We propose a method that combines finite state machines with fuzzy logic to represent concepts such as fast or slow which its definition may vary from agent to agent.