INFORMS 2021 Program Book

INFORMS Anaheim 2021

MC35

Operations of the fleet of shared autonomous vehicles (SAVs) can lead to an increase in empty trips such as repositioning trips with negative impacts on overall traffic and on the SAV fleet operator’s profit. To reduce inefficiencies of such services, this study proposes a personalized incentive scheme which is innovatively established based on a win-win strategy between households and the operator. Due to computational complexity, the scheme is iteratively developed upon an integrated framework consisting of the travel behavior model of households and an SAV fleet operations model. The results of the empirical investigation show that the framework offers the service provider an effective incentive scheme resulting in higher profit and more efficient operations of the fleet and presents travelers more SAV rides within the same budget. 2 - Service Bundle Design and Pricing for Shared Autonomous Vehicles (SAV) Qingyang (Tom) Xiao, SUNY at Buffalo, Buffalo, NY, 14228-3226, United States With the adoption of Autonomous Vehicles (AVs) and the rise of the concept of Mobility-as-a-Service (MaaS), we tackle the service bundles design and pricing problem. Through service bundles, travelers receive services at lower cost compared to Pay-Per-Ride scheme and in return, operators secure more demand. We formulate a two-stage hierarchical optimization model with the first stage problem of designing and pricing service bundles and second stage of minimizing operations for the service. A case study based on forecasting demand data for New York City (NYC) by New York Metropolitan Transportation Council (NYMTC) is conducted to derive insights of this system. 3 - Adopting Automated Vehicles and Creating Equitable Transport Markets Automated vehicles (AV) provide opportunities to improve public transportation (PT) in low-density suburbs. However, AVs may adversely affect PT demand and service. In this research, we aim to maximize positive AV externalities such as under-served population mobility whilst minimizing negative externalities like congestion. We develop a stylized model connecting a suburb to a central business district and compare outcomes over multiple market structures. We assess monopolistic to competitive AV markets and AV-PT cooperation for a single representative period and peak / off-peak periods, taking into account AV fixed costs. We compare the market structure results to an optimal social welfare outcome and derive subsidies and congestion charges. The findings may aid policymakers to develop regulatory tools which generate positive transport equilibria outcomes. 4 - Mixed Employment Modes for On-demand Platforms Facing Worker Benefit Ziwei Cao, University of Maryland-College Park, College Park, MD, 20740, United States, Michael O. Ball New regulations for gig-economy workers seek to convert independent contractors to employees. In this paper, we consider the pricing and staffing strategies of an on-demand ride-sharing platform adopting the mixed employment mode, i.e., hiring both part-time and full-time drivers, under mandatory benefit rules for full-time drivers. In our analytical model, drivers may decide whether to work for the platform and if so, whether to choose a full- or part-time work schedule. Our results, based on both analytical and numerical studies, show the impact on worker compensation and company profits under a variety of assumptions. Insights into potential government policies are also provided. MC37 CC Room 210C In Person: Telecommunications and Network Analytics 1 General Session Chair: Austin Buchanan, Oklahoma State University, Stillwater, OK, 74078-5017, United States 1 - Heuristics for the Budget-Constrained Immobile Server Problem Adam Quentin Colley, Southern Methodist University, Florence, AL, 35630-2617, United States Given a set of Poisson traffic streams (customers) and a fixed budget for opening and provisioning M/M/1 service queues at a set of potential locations, the Budget- Constrained Immobile Server Problem (BCISP) is to determine the number, location, and service capacities of the queues, and an assignment of customers to the queues that minimizes a cost function comprising fixed queue-setup costs and variable costs for customer assignment and waiting time. We propose heuristics for the BCISP that are easy and inexpensive to implement, and compare their performance against exact methods implemented with commercial mathematical programming software. Amir Brudner, The Hebrew University of Jerusalem, Mount Scopus Campus, Jerusalem, 9190501, Israel, Nicole Adler

MC35 CC Room 210A In Person: Intersection Control with Connected and Autonomous Vehicles General Session Chair: Mojtaba Abdolmaleki, University of Michigan, Ann Arbor, MI, 48109, United States 1 - Stochastic Reservations for Autonomous Intersection Management Carlin Liao, University of Texas at Austin, Austin, TX, 78705-3030, United States, Stephen D. Boyles Automated intersection management relies on sequencing movements with tight clearances to increase throughput compared to traffic signals, but this assumes that all vehicles are automated and precise. To accommodate human drivers, the intersection must increase the space reserved for them at the cost of efficiency. Since these vehicles can’t use the entire area allocated to them, this presentation will introduce the concept of stochastic reservations to allocate one timespace unit to multiple vehicles to be both safe and efficient, informed by driving simulator experiments with real human drivers. 2 - Performance Evaluation of Modified Cyclic Max Pressure Controlled Intersections in Realistic Corridors Simanta Barman, University of Minnesota, Minneapolis, MN, United States, Michael W. Levin Max pressure is an actuated decentralized signal control policy, proven to be stable for any stabilizable demand. This study focuses on evaluating the performance of this policy compared to the current signal control. Simulation models of seven intersections comprising two corridors, County Road (CR) 30 and CR 109 from Hennepin county, Minnesota were created. Then a cyclic max pressure control with realistic first-in-first-out queueing behavior was implemented. Comparisons based on average waiting time, vehicle speed etc. were then made to determine the better policy. This study aims to demonstrate that max pressure performs better than current signal timing based on realistic simulations. 3 - A Unifying Framework for Intersection Control Based on Graph Coloring Mojtaba Abdolmaleki, Graduate Student Research Assistant, University of Michigan, Ann Arbor, MI, United States, Yafeng Yin, Neda Masoud This talk discusses a unifying mathematical framework for intersection control that aims to allocate right-of-way to conflicting traffic movements to minimize delay or maximize throughput. Given the problem’s NP-hardness, we devise a hybrid LP-relaxation graph coloring approximation algorithm to find the optimal control for a generic demand pattern. Assuming the footprint of the intersection is sufficiently large, we prove the algorithm is a polynomial-time approximation scheme. Our scheme is unifying in the sense that it includes as a special case the traditional signal control for manually driven vehicles and the reservation-based signal-free schemes for automated vehicles. 4 - A Multiclass Link Transmission Model for Dynamic Network Loading of Mixed Legacy and Automated Vehicle Flow Michael W. Levin, University of Minnesota, Minneapolis, MN, United States, Di Kang Many cities will experience a mixed traffic flow consisting of both legacy and automated vehicles. Although the overall market penetration may be known, the proportion of automated vehicles may vary in space and time. Since automated vehicles are expected to behave differently than legacy vehicles, this results in a flow-density relationship that varies in both time and space with the local proportion of automated vehicles. We model this scenario using a multiclass kinematic wave theory. We develop a multiclass Newell’s method for finding exact solutions to the multiclass kinematic wave theory. We then extend this method to a multiclass link transmission model. Numerical results from dynamic traffic assignment on the downtown Austin network demonstrate the computational tractability of this method and explore the effects of automated vehicles on traffic congestion. MC36 CC Room 210B In Person: Future Mobility and Urban Community General Session Chair: Ziwei Cao, University of Maryland, College Park, MD, 20740, United States 1 - An Integrated Personalized Incentive Scheme for Shared Autonomous Vehicles Somayeh Dejbord, University at Buffalo, Tonawanda, NY, 14150- 2856, United States

68