Reinforcement learning for Pedestrian evacuation Simulation and Optimization during Pandemic and Panic situation

Abstract

The pandemic of Covid-19 has drawn significant attention of the people around the world. The current situation has revealed that this virus infected more than 50 million people globally. On the other hand, the study on crowd simulation can demonstrate the behaviour of massive people that gather in the same location. The importance of this study can lead to safe evacuation in case an outbreak happens. This paper proposed a solution of innovative crowd simulation supported by reinforcement learning and pandemic factors. The study has successfully demonstrated the entering and leaving the various exit and entrance. The experiment is performed with a different mode such as two elevators for exit and entry, staircase and six lanes of gates. As a result of the experimental study, it reveals around 87% of agent behaviour has similarity compared to a real-life simulation performed by previous research. It means the simulation is reflecting the real-life human behaviour when the evacuation process has occurred. The future study can extend to a contagion model of crowd behaviour where an agent can influence each other in a particular situation.