Crowdsim : a framework to estimate safety of egress performance in real life scenarios

Abstract

The power of the technology impacts on the modern life in different manners everyday. A variety of scientific models and computational tools have been developed to improve human safety and comfort in built environments. In particular, the comprehension of pedestrian behaviors has higher importance in such context. Moreover, some places are propitious to people agglomeration, i.e. a train station and an airport or even specials events like a sport game. In this thesis we discuss the use of crowd simulation to reproduce and evaluate egress performance in specific scenarios. We present CrowdSim: a crowd simulation framework designed to automatically reproduce crowd behaviors during an egress process. A set of software validation tests has been conducted in order to ensure the accuracy of the simulations computed by CrowdSim. In addition, CrowdSim is able to generate a set of plausible egress plans to be performed in a specific environment and also to rank them. Several case studies were performed in order to evaluate the work. Further, we performed different analyzes, including also the comparison of simulation data against data obtained from real-life experiments.