电影天堂

湖畔问道·鼎新论坛|Dynamic Bid-Price Strategies in Hub-and-Spoke Networks: Structural Properties and a Theory-Informed AI Agent

发布时间:2026-11-09浏览次数:10

讲座题目

Dynamic Bid-Price Strategies in Hub-and-Spoke Networks: Structural Properties and a Theory-Informed AI Agent

主讲人

(单位)

冯有翼

西班牙萨拉戈萨物流中心(Zaragoza Logistics Center)

主持人

(单位)

陈桢

(电影天堂 )

讲座时间

2026年11月12日下午2:30

讲座地点

综合楼113

主讲人简介

冯有翼(教授),男,于南京大学获数学专业学士及硕士学位,以及于美国哥伦比亚大学获工业工程专业硕士及博士学位。曾任职于纽约Bearing证券、美国休斯顿安然公司、新加坡国立大学计算机学院信息系统系、香港中文大学系统工程与工程管理系、西班牙萨拉戈萨物流中心(Zaragoza Logistics Center)等任教多年。研究方向为供应链管理、库存管理、收益与风险管理。其研究成果发表于《Operations Research》、《Management Science》、《IEEE Transactions on Automatic Control》、《IIE Transaction》、《European Journal of Operational Research 》、《Naval Research Logistics》等运筹与管理科学领域国际权威期刊。

讲座内容摘要

We address dynamic bid-price strategies for a capacity-constrained hub-and-spoke network where n spoke cities feed into a single hub. This topology introduces complex inventory interactions involving both complementarity (connecting flights) and competition (local vs. connecting traffic). We prove that the optimal value function exhibits coordinate concavity alongside mixed submodularity and supermodularity. These fundamental structural properties rigorously characterize the bid prices---the underlying opportunity costs of aircraft seats---across the entire network. Leveraging these theoretical guarantees, we propose a“Theory-Informed AI”framework for an automated revenue management AI agent. This approach employs constrained deep learning to estimate unknown demand intensities while strictly enforcing the proven geometric properties of the bid prices, effectively mitigating the curse of dimensionality and delivering robust, stable pricing policies for high-dimensional networks.