Improving Air Mobility in Emergency Situations

Abstract

Emergency situations in aviation pose serious risks to life and result in huge negative impacts on air mobility, causing a significant economic and reputation loss to airlines and airports. However, the decisions to deal with emergencies are usually made by flight dispatchers according to their experience, and they merely consider local-view optimization. Therefore, there is an urgent need to design a decision-making assistant system to alleviate the negative impact of perturbations on aviation air mobility in the global-view perspective. In this project, we will develop a framework based on machine learning that captures the patterns of emergency situations and optimizes the operation schedules quickly and accurately for maximum air mobility efficiency at both micro-level and macro-level. We will utilize multi-source data and leverage deep learning models to predict the consequence of emergency events considering the spatial-temporal characteristics of the events. Based on a prediction model, we will optimize air mobility output by adopting a deep multi-agent reinforcement learning model. Our goal is to provide prealert and decision-aid system for passengers and airport staff when emergency events occur, and to adjust the original schedule for quick recovery of disrupted air mobility

Yongxin (Jack)​ Liu - ERAU