Application of artificial neural networks for agent-based simulation of emergency evacuation from buildings for various purpose

Abstract

The application of Artificial Neural Networks (ANN) for the pedestrian flow simulation is a new stage in the development of system simulation, which has become accessible due to the exponential growth of computing power. Authors, together with colleagues from the Beihang University, Beijing, developed a program that allows to solve practical problems connected with emergency evacuation in construction using system simulation based on ANN. Machine learning allows us to precisely simulate the behavior of all people during evacuation, their reaction on obstacles and other people, and predict the load on the main evacuation routes in advance and, as a consequence, make changes to the plans of the buildings, where necessary. With the help of this program architects and designers can find the right solutions for the projects, which will not require any further adjustments, after the building will be constructed. This program is especially important for unique multifunctional facilities, for example, sports stadiums or shopping and entertainment centers, where a real evacuation check cannot be carried out.