Pedestrian Movement Simulation in Evacuation Process from a Dynamic Environment using Agent-Based Modeling

Studying pedestrian movement behavior is a necessary tool before the construction design in order to predict the social and collective behavior in different situations. Considering the complexity of human behavior using simulation is indeed a significant step for managers and designers in different fields including transportation planning. Agent-based modeling is one of the most popular techniques in this area. This paper aims to present a novel hierarchical agent-based simulation of pedestrian evacuation from a dynamic environment using reinforcement learning which is the closest to human behavior among the other machine learning algorithms. In the approach agents autonomously decide through a three-layer hierarchical model which includes goal, node, and cell selection layers. All the approach was originally coded in NetLogo software. The proposed model was successfully applied to simulate the pedestrian evacuation process from the platforms of the Shahed Metro station in Tehran during a destructive event. The results showed that practical capacity of the evacuation from the metro platform is 1.9 and 3.4 in the case of 2 and 4 exit doors respectively. The results from the approach can be used by building designers (non-structural design) and managers to optimize the quality of evacuation; also the proposed model has the potential of being used to analyze the evacuation time in the case of emergency (fire, earth-quake, terrorist attack and etc.)