Reinforcement Learning-based Unpredictable Emergency Events

Abstract

The vehicle routing problems is one of the wildly known transportation problems. It used to minimize the total traveling time of vehicles by choosing the shortest path. Defining the routing of the vehicles in the real world is a complex task to perform because of the different constraints to handle. The aim of this paper is to develop a dynamic simulation environment using Java for testing Q-learning approach with consideration of overall and dynamic performance. We propose Q-learning based approach in order to improve the transportation facilities for emergency response activity. With the aim of minimizing the time from emergency call being waited for a relief or a service to the dispatch point. The results showed that optimisation scheme, developed by the RL agents based on Q-learning approach using simulated environment, has the potential to offer an accurate scheme to find the optimum route.