There has been a high incidence of sudden-onset natural disasters worldwide in the recent years. The dispatch of emergency supplies is a key aspect of post disaster relief. Many natural disasters induce secondary disasters and other damages. During rescue operations, roads are often found to be damaged and impassable; nonetheless, secondary disasters often cause the demand for emergency supplies to evolve dynamically. Previous studies have mostly focused on static vehicle-routing problems, which fail to simulate the dynamic changes at the affected sites, realistically. This study considers the dynamic distribution of emergency supplies using a hybrid fleet of trucks and drones. A mixed-integer programming model is constructed by allowing the carrier vehicles to depart from the distribution center multiple times with the dual objectives of minimizing both the road risk and total waiting time for supplies at the affected sites. Because this problem is an NP-hard problem, this study designs a biobjective optimization algorithm that incorporates the nondominated sorting genetic algorithm-II (NSGA-II) and Q-learning (QLNSGA-II). The results of computational experiments show that, for small-scale instances, the solutions of QLNSGA-II are very close to the exact solutions obtained using the ϵ-constraint method. For large-scale instances, the solutions of QLNSGA-II outperform those of the NSGA-II algorithm and the hybrid algorithm of NSGA-II and random neighborhood search in some performance metrics, such as the hypervolume and inverted generational distance. Finally, this paper presents a geological disaster event that occurred in Zigui County, Hubei Province, in China, as a case study to show how the proposed model works. This study offers a practical framework to guide decision-making while scheduling emergency-supply deliveries.
扫码关注我们
求助内容:
应助结果提醒方式:
