In this work with NYUSH professor Zhibin Chen and his PhD student Yuhao Liu, we consider resource allocation problems where the preferences of users in the system are not known in advance. The topic involves designing quarantine hotel allocations during COVID but the applications can extend to ticket sales, humanitarian logistics, and cybersecurity.
The work can be accessed here.
The latest polls from Shanghai Rankings for the subject of Transportation Science & Logistics shows NYU in the top ten among US academic institutions. The ranking is related to research, which is a testament to the quality of our graduate programs in attracting top students to contribute to this field.
The lab had our second Christmas book exchange (the first was in 2023), which was a great success!
To aid policymakers in evaluating the impact of a project, our collaborative team at NYU investigated new measures that better quantify project outcomes on different population segments, using Citi Bike as a case study.
Paper available here.
In this work with Jelly Ziyue Li and Qianwen (Vivian) Guo at FAMU-FSU College of Engineering, we explored the use of analytical real options models to manage capacity expansion of airport runways under uncertainty. The proposed system modeling framework includes “jump diffusion processes” that account for disruptions in system capacities. While both conventional reinforcement learning and analytical real options models deal with optimal policies, the latter assumes underlying variables follow simple stochastic processes, ie random distributions over time, and derive optimal threshold policies (e.g. if demand variable exceeds threshold H, execute decision). The approach makes it easy to quantify the impacts of uncertainty on timing of strategic decisions. (e.g. disruption rates increased 10%? that may be reflected by an increase in the value of having the option to make the decision in the future).
Paper link: https://www.sciencedirect.com/science/article/abs/pii/S0969699725000870?dgcid=coauthor
In this work with Hai (Marshall) Yang, and Purdue University colleagues Tho Le, and Yuchen Du, we developed a highly scalable algorithm for designing fleet and battery sizing for sidewalk delivery robots like those of Starship Technologies and Kiwibot. Test results make use of synthesized data based on the Purdue campus setting, and are available to share with other researchers. Funding from National Science Foundation (NSF). Department of Civil & Urban Engineering at NYU TandonC2SMARTER Center
Paper link: https://www.sciencedirect.com/science/article/abs/pii/S1366554525002613?dgcid=author
In this work with Xiyuan Ren and Prateek Bansal, we examine mixed logit market share models with market-specific parameters and develop a method to estimate them. We leverage data shared by Replica to show that roughly 40% of the NYC population value driving into lower Manhattan at $9 or more, whereas 90% of travelers from outside NYC into lower Manhattan value driving in the same way. The method was used to produce the NYS statewide mode choice model parameters which are freely accessible on Zenodo (https://lnkd.in/duqn6p9w).
Paper link here.
In this collab with KAIST (Yichan An, Jinwoo Lee) and Konkuk University researchers (Soomin Woo), we looked at the large-scale spacing and sizing design of electric charging stations in a region with spatial variations in demand and travel time, taking into account the delays from waiting for charger availability. An analysis of the five boroughs of NYC illustrates its applicability and improvement over a naive assumption without such variations. This provides a new tool for policymakers to plan resources for charging infrastructure at a much larger scale than before.
Paper available here.
Two of the challenges for public agencies evaluating privately-operated Mobility-on-Demand (MOD) systems: capacities are dynamic, but the operational policies are private (unobservable). We (Bingqing Liu, David Watling) invented a new method to evaluate the steady state equilibrium performance of these dynamics that can be fitted to data using inverse optimization, without needing to assume knowledge of the company’s operational policy. This breakthrough allows agencies to evaluate impacts of changes in demand, supply, or operating policies, to different multimodal MOD systems. Builds off earlier work developed by JIA XU. Pre-proof in EJOR in link below.
https://www.sciencedirect.com/science/article/pii/S0377221724009809
What are the large-scale fleet sizing ramifications for sidewalk delivery robots? Hai (Marshall) Yang and Prof. Chow collaborated with Tho Le and Yuchen Du (Purdue University) to investigate this problem. We developed a new analytical model of pickup and delivery problems with bounded makespans to study such fleets. Using neighborhoods in NYC as case study, we observed: a diminishing effect of economies of scale when demand increases; a high dependency on dwell/service time at each stop on the operating cost; and the major impact of sidewalk level of service. Paper link:
https://www.sciencedirect.com/science/article/pii/S0968090X24004996