SIMULATION OF HUMAN CROWD BEHAVIOR BASED ON INTELLECTUAL DYNAMICS OF INTERACTING AGENTS

This paper elaborates a phenomenological approach to simulation of human crowd behavior, proposed in study [1]. We consider a continuous stochastic agent-based model of human behavior in a confined space with a given geometry by using refinements of both an agent status and agent's decision-making system, presented in Helbing's models [2, 3, 4] (molecular approach). Such integration seems to be the most promising development of this class of tasks due to the fact that the phenomenological approach (Beklaryan-Akopov's model) enables to introduce natural discretization of a task and then to calculate the increment of all agent's characteristics at any specific time, and the use of elements of the molecular approach (Helbing's model) enables to describe the most realistic decision-making system of an agent. This removes a complicated issue of numerical integration of Newton's equations underlying Helbing's model and offers explicit calculations of all system characteristics. As a result, an agent based model has been devised in AnyLogic simulation modeling system, enabling to investigate agent movement dynamics with due regard to "the crowd effect" in various scenarios, in particular, in extreme situations, when exposure to "crowd crush" and "turbulence" effects exists.