1. Introduction
Simulating pedestrian evacuation is a very important topic which involves different disciplines such as social psychology, crowd engineering and computer simulation [1], [2]. Different simulation models including flow-based, cellular automata, and agent-based have been proposed [2]. The flow-based modeling such as EVACNET4 enables the user to construct a simulated physical environment as a network of nodes, provide the specification of arcs, determine how many people the particular node may contain, and supply the traversal time and flow capacity for each arc [2]. This kind of modeling can be considered as the macroscopical modeling for the pedestrian evacuation. However, the flow-based modeling is hard to simulate the crowd behavior emerged from the interactions among pedestrians. On the other hand, the cellular automata modeling and agent-based modeling can be considered as the microscopical modeling for the pedestrian evacuation. The main difference between the cellular automata modeling and other modeling types involves the discretization of space [30]–[33]. In the cellular automata modeling, the space is discretized into cells and each cell is either occupied by a pedestrian or an obstacle or is empty. In each time step, the pedestrian moves from cell to cell on the basis of the throw of weighted die. The cellular automata modeling is suitable for large-scale evacuation simulations. However, this type of modeling is not able to reproduce all collective effects and self-organization phenomena observed empirically. The well-known representative of agent-based modeling is the Social Force Model proposed by Helbing and Molnar [8], which treats pedestrians as particles subject to the long-ranged forces induced by the social behavior of the individuals.