Q-Learning Based Optimal Escape Route Decision in a Disaster Environment

It is important to evacuate quickly to a safe route since most of the deaths occur in the initial fire. However, there is a problem that a person in a building cannot decide considering other floors situation. Therefore, a system is needed to reduce the damage caused by fire by presenting an optimal escape route considering the situations of other floors. Thus, we propose a Q-Learning based system model to find the optimal escape route from the current location of the user, considering the fire in a building. The shortest path is predicted according to multi-story building structure and considers the different location of exits and fire each floor. Also, consider that if there are many people around the current location of the user, selecting only the shortest path may cause bottlenecks and poor evacuation. We propose a system model that appropriately distributes the optimal escape route and the next optimal escape route according to the number of people around the user and confirm the results through simulation.