Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TA30

multiple travelers on network route resources. Furthermore, we design an adaptive centroid-based clustering algorithm to find a local optimal clustering solution. Built upon that, a mixed strategy coordinated routing mechanism is implemented to coordinate in-vehicle routing decisions for multiple traveler clusters over a large scale network. The numerical experiments are conducted to validate the efficiency and applicability of the proposed approaches. 2 - Macroscopic Traffic Flow Modeling with Mixed Connected and Human-driven Vehicles Xianfeng (Terry) Yang, University of Utah, Salt Lake City, UT, United States, Zhehao Zhang, Zhao Zhang Although connected vehicles (CVs) will soon go beyond testbeds, it can be expected that CVs and human-driven vehicles (HVs) will co-exist over a long period. Due to the capability of exchanging data, CVs behave differently compared with HVs. In this study, we aim to develop a macroscopic traffic flow model to understand how speed change of CVs would impact HVs in the traffic stream. Particularly, friction factors are introduced to the speed formulations for accounting interactions between CVs and HVs. Then extended Kalman Filter is employed to update both model parameters and friction factors in real-time. For model evaluation, we employ simulated data as ground truth for conducting numerical tests. 3 - Information Dissemination Dynamics via Vehicle-to-vehicle Communication over Transportation Networks Ala S. Alobeidyeen, University of Florida, Gainesville, FL, United States The study developed a discrete mathematical model to track information spreading dynamics via V2V over an urban transportation network at discrete time steps. Specifically, an information network flow combined with the IT-CTM (Du et al., 2016) is developed to respectively track traffic information wave spreading dynamics at traffic intersections and road segments. Moreover, machine learning approach is used to investigate the correlation between information coverage and traffic congestion based on simulation experiments. 4 - Online Demand-driven Car Sharing Rebalancing Xiaopeng Li, University of South Florida, 4202 E. Fowler Avenue, Tampa, FL, United States, Dongfang Zhao This study proposes an online car-sharing rebalancing model to deal with the car sharing fleet management problem using reinforcement learning. We develop a multi-agent reinforcement learning framework using a deep Q-learning algorithm. The goal of the algorithm is to maximize the total profit of the platform by repositioning available cars to the locations with outstanding demand-supply gaps. This model does not make any assumption on the demand and is completely driven by spatially distributed demand data history. With real-world data, we show significant improvements of the proposed framework over state-of-the-art approaches through extensive empirical studies. Chair: Niels Agatz, Erasmus University, Rotterdam, Netherlandsl 1 - Splitting Pickup and Delivery Tasks in a Same-day Personal Shopper Service Alp Arslan, Erasmus University Rotterdam, Admiraliteitskade 50-713, Rotterdam, 3062ED, Netherlands, Niels Agatz, Mathias A. Klapp We consider a same-day personal shopper service that receives customer delivery requests that require pickups at one or more stores. All requests that can be served in their requested time window are accepted. We study the benefits of splitting a single customer request into various delivery tasks served by different shoppers to improve the percentage of customers served throughout the day. 2 - Tactical Design of Same-day Delivery Systems Alex Stroh, Georgia Institute of Technology, Atlanta, GA, 30332, United States, Alan Erera, Alejandro Toriello We study tactical designs for same-day delivery (SDD) systems. We look for structure in optimal dispatch policies when a SDD system is modeled with smoothed customer demand and vehicle route durations are approximated by a continuous time function of the number of customers served. Using these optimal policies, we study tactical decisions for SDD systems which prior research has used as model inputs. Specifically, we can answer questions like: How large of a fleet is required? When during a service day should a retailer stop accepting same-day orders? How often can a retailer expect to dispatch delivery vehicle? We illustrate our findings with a set of realistic examples. n TA32 North Bldg 222B Same-day Delivery Sponsored: TSL/Freight Transportation & Logistics Sponsored Session

n TA30 North Bldg 221C Transportation Aspects Towards a Smart City Emerging Topic: Smart Cities

Emerging Topic Session Chair: Stephen Mattingly

1 - The Marginal Congestion of a Taxi in New York City Alejandro Molnar, Asst Professor, Vanderbilt University, 415 Calhoun Hall, Nashville, TN, 37240, United States, Daniel Mangrum We exploit the partial deregulation of New York City taxi medallions to provide a causal estimate of the impact of taxi supply on congestion. We employ taxi trip records to measure historical street-level speed. We find that the roll-out of newly authorized taxis caused a local 8-9% decrease in speed. We estimate an empirical congestion elasticity curve from heterogeneous changes in speed and taxi supply, counted from aerial orthoimagery. Additionally, we provide novel urban sensor data to document a substantial traffic slowdown since 2013. Most of the slowdown in midtown Manhattan is accounted for by new supply from ridehail applications. 2 - Integrating Transportation Education and Research for Smart Cities Ruey Cheu, University of Texas at El Paso, El Paso, TX, United States, Miroslav Svitek, Tomas Horak The concept of smart cities requires professionals in different disciplines to work together and engaging the community, but there is a lack of such education program. In this presentation, the speakers will examine the strength and weaknesses of discipline specific transportation education programs in universities. They will then discuss several smart cities education and research initiatives at The University of Texas at El Paso (UTEP) and Czech Technical University (CTU) at undergraduate, graduate and PhD levels. They will next share their experience in transforming the dual master degree program in transportation between UTEP and CTU, to a new dual master degree program in smart cities. 3 - Decentralized Optimization of Urban Traffic Flows Stephen F. Smith, Carnegie Mellon University, Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, 15213, United States Recent work in decentralized, schedule-driven traffic signal control has produced a promising new approach to traffic flow optimization in urban road networks. This approach, which relies on a novel formulation of the intersection control problem as a special type of single machine scheduling problem, has yielded significant efficiency improvements in initial field deployments. In this talk, we describe current research aimed at broadening network-level optimization performance through use of queuing theory concepts, more expressive problem representations, and bi-directional information exchange. 4 - I-STREET Testbed: Progress and Challenges Clark Letter, University of Florida, Gainesville, FL, United States, Lily Elefteriadou The University of Florida and its Transportation Institute (UFTI), the Florida Department of Transportation and the City of Gainesville have coordinated the development of a smart testbed on the UF campus and adjoining city streets. The testbed is established to deploy and evaluate numerous advanced technologies including connected and autonomous vehicles, smart devices, and sensors, as well as to develop novel applications for their use. Many challenges arise when deploying new technologies, and I-STREET hopes to share lessons learned with other agencies. n TA31 North Bldg 222A Integration of Learning Approaches and Emerging Traffic Operation and Control Technologies Sponsored: TSL/Intelligent Transportation Systems (ITS) Sponsored Session Chair: Lili Du, University of Florida, 2550 Yeager Road, Gainesville, FL, United States 1 - Clustering Based Mixed Strategy Coordinated In-vehicle Routing for Connected and/or Autonomous Vehicles Wang Peng, University of Florida, Gainesville, FL, United States This study develops a clustering based mixed strategy coordinated in-vehicle routing approach. Specifically, this study develops graphical and network flow approaches to evaluate the direct and indirect competition potential among

254