摘要
针对地震等极端灾害场景下地面道路损毁导致车辆难以抵达且无人机受续航等限制无法长距离直达受灾点的困境,提出了建立车辆与无人机协同运输的多级应急物流网络。在该网络框架下,为了保障运输效率以及有效控制成本,构建了兼顾时间与成本的无人机升降点选址多目标优化模型。为了求解该模型,设计了一种数据驱动的遗传算法(Data-Driven Genetic Algorithm, DDGA)。针对传统算法依赖人工手动调参的局限性,该算法通过分析寻优过程中累积的历史试验数据,构建超参数与性能的映射模型,实现对种群规模、交叉率及变异率等关键超参数的自动寻优。通过不同规模算例的数值实验进行分层验证:在小规模算例中,比较模型求解结果与商业求解器Gurobi的解,验证了模型的正确性;在大规模算例中,商业求解器无法在5 400 s内计算出最优解,而DDGA算法能够获得高质量可行解,验证了算法求解大规模问题的高效性与准确性。以某次真实地震场景为背景进行算例分析,获得了应急物流网络总成本为287.79万元、应急救援总时间为404.76 h的升降站选址方案,有效缓解了受灾地区应急物资运输的时效滞后与成本压力。根据研究结果,所构建的车辆与无人机协同运输的多级应急物流网络更加适配极端灾害场景,基于超参数自动寻优的数据驱动遗传算法克服了人工调参试错的低效性与不确定性,突破了大规模选址问题的求解瓶颈,从而为受灾地区无人机升降站布局提供了兼顾时间与成本的科学方案。该研究成果可为无人机在应急物流中的应用提供参考。
Addressing the dilemma that the damage of ground roads in extreme disaster scenarios such as earthquakes makes it difficult for vehicles to reach and UAVs cannot reach the disaster site for a long distance due to restrictions such as endurance,a multi-level emergency logistics network for coordinated transportation of vehicles and UAVs was proposed.Under the framework of the network,in order to ensure the transportation efficiency and effectively control the cost,a multi-objective optimization model of UAV take-off and landing site selection was established,balancing time and cost.To solve the model,a Data-Driven Genetic Algorithm(DDGA)was designed.In view of the limitations of traditional algorithms relying on manual parameter adjustment,this algorithm constructed a mapping model between hyper parameters and performance by analyzing historical trial data accumulated during the optimization process,achieving automatic optimization of key hyper parameters such as population size,crossover rate,and mutation rate.This study conducted tiered validation through numerical experiments of varying scale examples.In small-scale examples,the model′s solution was compared with that of the commercial solver Gurobi,validating the model′s correctness;in large-scale examples,Gurobi failed to calculate the optimal solution within 5400 seconds,while the DDGA algorithm obtained high-quality feasible solutions,validating its efficiency and accuracy in solving large-scale problems.Through an example analysis based on a real earthquake scenario,a landing site selection scheme with a total emergency logistics cost of CNY 2.8779 million and a total emergency response time of 404.76 hours was obtained,which effectively alleviated the time lag and cost pressure of emergency material transportation in disaster areas.According to the research results,the multi-level emergency logistics network of vehicle and UAV coordinated transportation is better adapted to extreme disaster scenarios.The DDGA based on hyper parameter automatic optimization overcomes the inefficiency and uncertainty of manual trial-and-error parameter tuning,and breaks through the computational bottleneck of solving the large-scale site selection problem,and thus providing a scientific scheme for the layout of UAV landing sites in disaster areas,which takes into account both time and cost.The research results can provide reference for the application of UAV in emergency logistics.
作者
陆后军
贺博强
高银萍
LU Houjun;HE Boqiang;GAO Yinping(College of Logistics Engineering,Shanghai Maritime University,Shanghai 201306,China;College of Transport and Communications,Shanghai Maritime University,Shanghai 201306,China)
出处
《交通运输研究》
2025年第6期151-164,共14页
Transport Research
基金
国家自然科学基金项目(72501171)。
关键词
应急物流
无人机
升降站点
选址规划
数据驱动
遗传算法
emergency logistics
UAV(Unmanned Aerial Vehicle)
landing site
site selection planning
data driven
genetic algorithm
