Dynamic Cellular Learning Automata for Evacuation Simulation

Pedestrian behaviors arc essential for the evolution of an evacuation process and the eventual evacuation time. Such behaviors are mainly determined by the information available to a pedestrian at a certain position. To obtain a more comprehensive base for sensible movement choices, a pedestrian is expected to extend the range of information acquirement, which can be reflected in spatial and temporal concepts. The existing models for evacuation simulations mainly address the extension of information in a spatial concept. In fact, a pedestrian is more likely to update his movement choice based on the experiences accumulated throughout the evacuation process, as defined by information extension in a temporal concept. This paper develops a model to achieve an adaptive choice-making process. Two learning automata are proposed to update the imitation preferences for certain pedestrians and to make decisions between independent movement and imitation. Sensitivity analyses are performed using several key parameters to understand the mechanism of the proposed model. The proposed model is compared with the floor field model in terms of evacuation time and process, showing a high performance in describing pedestrian movement characteristics at different stages of an evacuation process.