INFORMS 2021 Program Book

INFORMS Anaheim 2021

SD37

platforms. To study this, we conduct a series of experiments in a restaurant food delivery setting. Our results show that consumers indeed tend to choose the platform where their chosen restaurant is rated higher, even when they know they will receive exactly the same service. This may imply that in face of competition a platform may find it disadvantageous to counter vendors’ rating inflation. 2 - Dine in or Take Out? Trends on Restaurant Service Demand Amid the Covid-19 Pandemic Linxuan Shi, The George Washington University, Washington, DC, United States, Zhengtian Xu The outbreak of COVID-19 pandemic has caused unprecedented damage to restaurant dine-in services, given the concerns of exposure to coronavirus. In contrast, online food ordering and delivery services, represented by DoorDash, Grubhub, and Uber Eats, filled in the vacancy and achieved explosive growth. The restaurant industry is experiencing a drastic change under the crossfire of these two driving forces. However, due to the lack of first-hand data, we are not fully exposed to the underlying changes, let alone understand the potential impacts and launch targeted policies. To address such a pressing need, this study proposes to leverage the foot-traffic data to effectively keep track of the rapidly evolving demand for restaurant businesses. Data based on 0.8 million cellphone users and 10 thousand restaurants in the DC area is applied for demonstration and analysis. 3 - Drone Dispatch Policy to Fulfill Uncertain Customer Demands in a Delivery Network Zhenyu Zhou, Wayne State University, Detroit, MI, 48201-1111, United States We present a dynamic vehicle routing problem encountered in the design of an on-demand meal delivery network. Through subscription contracts each customer has the right to order a meal a day which will be delivered in, e.g., 20 minutes, by a drone. Customer locations are aggregated and represented by demand nodes in the network. In a delivery trip, a drone will start from a depot node, visit the demand node and return to (the same or a different) depot node. Not every demand node is reachable all depots. The drone dispatch is performed periodically, e.g., once every 10 minutes. The time slot in which a customer makes the order is uncertain. We present a stochastic dynamic programming model to maximize the total expected number of demands fulfilled by the end of the day. Reasonable state space reduction schemes will be presented to address the representation and computation challenges. SD37 CC Room 210C In Person: Vehicle Routing Contributed Session Chair: Xufei Liu, University of South Florida, Tampa, FL, 33613, United States 1 - Solution Approaches for the Rendezvous Vehicle Routing Problem Eric Oden, University of Maryland-College Park, College Park, MD, United States, Bruce L. Golden, S. Raghu Raghavan We consider a novel scheme for same-day delivery, in which a set of vehicles (shuttles) may intercept trucks moving along their fixed routes to transfer packages ordered at the last minute. This scheme can lead to significant transportation savings, as shuttles need not travel as far to accommodate the last- minute requests. We present a column generation algorithm which can quickly generate optimal solutions for reasonably-sized instances. We also develop and demonstrate the effectiveness of a specialized heuristic for use in larger instances. We then present results demonstrating the efficiency of truck-shuttle synchronization in various settings. 2 - A Column Generation Approach for a Stochastic Vehicle Routing Problem Eric Oden, University of Maryland-College Park, College Park, MD, United States, Bruce L. Golden, Subramanian Raghavan We consider a vehicle routing problem with stochastic travel times and service times. Furthermore, the customers are stochastic (i.e., each customer may cancel with some probability). We consider the problem of hiring trucks, assigning trucks to customers, routing trucks through their customers, and establishing appointment times, subject to fixed, travel, earliness/tardiness, and overtime costs. We present our column generation heuristic, as well as results concerning the subproblem of determining expected arrival times given the sources of stochasticity in the problem.

SD35 CC Room 210A In Person: Pricing in Shared Mobility Markets General Session Chair: Eduardo Marino, University of California, CA, United States 1 - Dynamic Simulation Model for Planning and Real-time Management of System of EV Fast-Charging Stations Dingtong Yang, University of California-Irvine, Irvine, CA, 92697, United States Motivated by the environmental benefits of and associated government regulations promoting electric vehicles (EVs), as well as the limited charging infrastructure to support EV travel in place, this study presents an agent-based stochastic dynamic modeling framework of a regional system of EV fast-charging stations to support the planning and real-time management of EV fast-charging stations. To model EV user fast-charging station choices, the framework incorporates a multinomial logit station choice model that considers station charging prices, expected wait times, and detour distances. Moreover, Each EV charging station is modeled as a multi-server queueing model. To manage the system of stations, this study proposes dynamic demand-responsive price adjustment (DDRPA) schemes based on station queue lengths. The computational results, based on a real-world system of EV charging stations in California, indicate that the best DDRPA scheme reduces average wait time by 26%, increases revenue by 5.8%, and increases social welfare by 2.7%. Moreover, the results illustrate how the modeling framework can identify stations that require additional chargers and areas that would benefit from additional fast-charging stations. 2 - Dynamic Parking Management for Automated Vehicles in Downtown Areas Tara Radvand, Graduate Student Research Assistant, University of Michigan, Ann Arbor, MI, United States, Sina Bahrami, Yafeng Yin This study proposes a dynamic model for the parking choice in a downtown area in the era of automated vehicles (AVs). Given the distribution of users’ activity time in the downtown, we propose a system of ordinary differential equations to model their AVs’ choice between an outskirt parking lot and cruising as a substitution for parking. Cruising may cause traffic congestion, which is captured by a network macroscopic fundamental diagram. With the proposed model, we further investigate dynamic time-based tolling strategies to optimize the system performance. 3 - Modeling Framework for Pricing-consistent Subscription Services in Shared Mobility Systems Eduardo Marino, University of California, Irvine, CA, United States, R. Jayakrishnan As shared mobility systems and various new paradigms of associated ownership and subscription systems are taking hold now, costs and prices in such systems need to be analyzed in depth. The current cost models are based on average values and long life-cycles, which are insufficient as daily travel miles of vehicles may significantly change. We present the conceptual aspects of the interactions of new cost structures and system performance in these new mobility systems. We present the properly designed cost function, a framework to analyze the interactions and optimize the new systems and provide results from an agent- based simulation of candidate contexts. SD36 CC Room 210B In Person: Emerging Topics in Food and Grocery Delivery Services General Session Chair: Qi Luo, Clemson University, Clemson, SC, 29634, United States Chair: Zhengtian Xu, The George Washington University, Ann Arbor, MI, 48105-2540, United States 1 - The Vertical Spillover Effect of Online Ratings on Platform Competition: An Empirical Investigation Yulia Vorotyntseva, Saint Louis University, Saint Louis, MO, 19102-4325, United States, Aleksi Aaltonen, Subodha Kumar, Paul Pavlou The familiar ‘five-star’ ratings system makes it easy for consumers to use product evaluations across competing platforms to choose a product or service. The average rating for the same product can vary across platforms for reasons unrelated to quality, including pure randomness. We argue that such diverging evaluations can give rise to a vertical spillover effect, that is, the evaluations of a product represented on a platform may affect the consumer’s choice between the

41