A Method of Delivering Fuel to Telecommunication Exchange Buildings in Disaster Response

In this paper, we propose a data-driven and end-to- end deep reinforcement learning-based method for delivering fuel to telecommunication exchange buildings right after large-scale disasters in order to restore or continue their services. Specifically, we generate a fuel delivery plan for telecommunication exchange buildings by proposing a complex neural network model and optimize the model with an end-to-end and data-driven based Actor-Critic method. This method accepts inputs of all features of telecommunication exchange buildings required for the disaster response from one end and outputs an optimized fuel delivery plan at the other end. The experimental results show the effectiveness, robustness, and functionality of our method both on a simulated dataset and a real corporation dataset with the information of 200+ real telecommunication exchange buildings. To the best of our knowledge, our work is the first to not only show potential practical usage in the disaster response of telecommunication services but also leverage the lack of real data or work records of past disaster response because of our proposed deep reinforcement learning-based optimization mechanism.