摘要
为解决目前民航服务模式难以满足旅客个性化、多元化动态需求的问题,聚焦旅客出行服务组合优化研究,构建超图刻画服务节点、协同关系及旅客需求权重矩阵,结合马尔科夫决策过程与深度确定性策略梯度算法,设计分层规划与优先经验回放优化机制。结果表明,优化算法较传统基线方法在所有场景中的优化效用归一化≥0.98,耗时下降至少62%;在20种场景中三类旅客算法耗时下降94.7%、94.7%、89.4%,40种场景算法耗时下降73.0%、61.1%、65.2%。结论表明,算法能保证求解精度与稳定性,显著提升效率、降低旅客耗时,为服务资源动态配置提供高效辅助决策。
To address the issue that the current civil aviation service model struggles to meet passengers′personalized and diversified dynamic needs,this study focuses on optimizing the combination of passenger travel services.Methodologically,a hypergraph is constructed to characterize service nodes,collaborative relationships,and the passenger demand weight matrix.By integrating the Markov Decision Process with the Deep Deterministic Policy Gradient(DDPG)algorithm,two optimization mechanisms—hierarchical planning and prioritized experience replay—are designed.Experimental results show that compared with the Dynamic and Value Approximation baseline methods,the proposed DDPG algorithms achieve a normalized optimization utility of no less than 0.98 in all scenarios,with a time consumption reduction of at least 62%.Moreover,the optimized DDPG reduces travel time by 94.7%,94.7%,and 89.4% for business,tourist,and economy passengers respectively under 20 service scenarios,and by 73.0%,61.1%,and 65.2% under 40 service scenarios.The conclusion indicates that the algorithm ensures solution accuracy and stability,significantly improves efficiency,and reduces passenger travel time,providing efficient decision support for the dynamic allocation of service resources.
作者
陈福荣
王靖琦
周钧
张凯伦
王旭
桂祺章
张皓瑜
CHEN Fu-rong;WANG Jing-qi;ZHOU Jun;ZHANG Kai-lun;WANG Xu;GUI Qi-zhang;ZHANG Hao-yu(Travelsky Technology Limited,Postal Code,Beijing 100000,China;Travelsky Mobile Technology Limited,Beijing 100000,China;Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China)
出处
《航空计算技术》
2026年第1期65-70,共6页
Aeronautical Computing Technique
基金
国家自然科学基金项目资助(52572358)
中国民用航空局民航旅客智慧出行重点实验室开放课题资助(ZHCX-2025003)。
