A Reinforcement Learning Model of Multiple UAVs for Transporting Emergency Relief Supplies

In large-scale disasters, such as earthquakes and tsunamis, quick sufficient transportation of emergency relief supplies is required. Logistics activities conducted to quickly provide appropriate aid (relief goods) people affected by disasters are known humanitarian logistics (HL), play an important role in terms saving the lives those affected. previous last-mile distribution HL, transported trucks helicopters, but these transport methods sometimes not feasible. Therefore, use unmanned aerial vehicles (UAVs) attracting attention due their convenience regardless disaster conditions. However, existing planning that utilizes UAVs may meet some requirements for post-disaster supplies. Equitable among shelters particularly a crisis situation, it has been major consideration study. this study proposes introducing three crucial performance metrics: (1) rapidity supplies, (2) urgency (3) equity supply amounts. We formulated routing problem multi-objective, multi-trip, multi-item, multi-UAV problem, optimize with Q-learning (QL), one reinforcement learning methods. performed multiple cases different rewards quantitatively evaluated each countermeasure comparing them. The results suggest model improved stability all evacuation centers when compared other models.