Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SA29

using a fleet of specialized vehicles (around 700 hundred school buses for Boston), we identify a multistage decomposition of the problem and propose integer programming formulations and efficient heuristics for each stage, improving on the state of the art. We also highlights interesting aspects of a collaboration with a public organization. This work allowed the city of Boston to save a yearly 4+ million dollars starting in 2017-18, reinvested in education. 3 - Stable Matching in Large Scale Ridesharing Systems Amir Tafreshian, University of Michigan Ann Arbor, 2350 Hayward St., 1430 Gg Brown Bldg., Civil And Environmental Engineering, Ann Arbor, MI, United States, Neda Masoud Ride-sharing systems require an effective and efficient optimization methodology to match drivers and riders in large-scale transportation networks. This matching problem can be modeled as an assignment game in a 2-sided market. The core of this game provides a set of stable matchings between riders and drivers. However, the solutions in this set are not equivalent from the perspective of different participants in the system. This paper proposes a framework to establish stable matches in a 2-sided market while considering a secondary objective function on the optimal payoffs. 4 - Dynamic Redistribution of Bike Sharing Systems Konstantina Mellou, Massachusetts Institute of Technology, Cambridge, MA, 02141, United States, Patrick Jaillet This talk focuses on the redistribution of bike sharing systems. Spatial and temporal imbalances of the user demand often lead to empty and full bike stations. Since reliable service is essential for the viability of these systems, operators use a fleet of vehicles to redistribute bikes across the network. We propose a mixed integer programming formulation for the dynamic version of the problem, and, combined with heuristics and decomposition techniques, we are able to solve large real-world instances. Since accurate estimation of the user demand is essential to plan for efficient rebalancing, we also develop approaches to consider lost and shifted demand, which very often are not taken into account. Joint Session TSL/ICS:Approximate Dynamic Programming and Reinforcement Learning for Routing I Sponsored: TSL/Freight Transportation & Logistics Sponsored Session Chair: Dirk C. Mattfeld, University of Braunschweig, Braunschweig, 38106, Germany 1 - Approximate Dynamic Programming for Planning Driverless Fleets of Electric Vehicles Lina Al-Kanj, Princeton University, Warren B. Powell The combination of a centrally managed fleet of driverless vehicles, along with the operating characteristics of electric vehicles, is creating a transformative new technology that offers significant cost savings. This problem involves a control problem for assigning requesters to cars, a planning problem for deciding on the fleet size and a pricing problem all of which are high dimensional stochastic dynamic programs. We use approximate dynamic programming to develop high- quality operational control strategies which are then used to determine the optimal fleet size. We also discuss surge pricing to smooth out daily peaks using We present the stochastic-dynamic inventory routing problem for bike sharing systems (SDIRP). The objective of the SDIRP is to avoid unsatisfied demand by dynamically relocating bikes during the day. To anticipate potential future demands in the current inventory decisions, we present a dynamic lookahead policy (DLA). The policy simulates future demand over a predefined horizon. Because the demand patterns change over the course of the day, the DLA horizons are time-dependent and autonomously parametrized by means of value function approximation, a method of approximate dynamic programming. We compare the DLA with conventional relocation strategies from the literature and lookahead policies with static horizons. Our study based on real-world data by the BSS of Minneapolis (Minnesota, USA) reveals the benefits of both anticipation by lookahead as well as the time-dependent horizons of the DLA. We additionally show how the DLA is able to autonomously adapt to the demand patterns. n SA32 North Bldg 222B an optimal learning approach in a real-time setting. 2 - Policy Search for Bike Sharing Systems Dirk C. Mattfeld, University of Braunschweig, Wirtschaftsinformatik, Braunschweig, 38106, Germany Jan Brinkmann, Marlin Ulmer

n SA29 North Bldg 221B

Joint Session TSL/Urb/Practice Curated: Activity Modeling and Applications in Urban Transport Sponsored: TSL/Urban Transportation Sponsored Session Chair: Jee Eun Kang, University at Buffalo, Buffalo, NY, 14260, United States 1 - Strategic Public Parking Locations with Routing Considerations of Individual Autonomous Vehicles Somayeh Dejbord, University at Buffalo, Tonawanda, NY, United States Use of personal Autonomous vehicles (AV) will lead to increased Vehicle-Miles- Traveled (VMT) particularly due to empty trips. While a decision maker makes the parking location decision (public), the decision of using the parking is made by individual private AV owners, households. A location siting model uses flow- based demand data as an input but the actual decision of using parking is made individually with finer details such as what household members’ daily itineraries are. We develop a methodology to approximate input data for the location model using data available to the decision maker. 2 - Integrating Augmented Reality, Gamification and Social Interactions in Mobile Apps to Influence Short- and Long-term Travel Decisions Yuntao Guo, Purdue University, West Lafayette, IN, United States, Srinivas Peeta, Shubham Agrawal, Irina Benedyk, Mohammad Miralinaghi This study develops a framework that integrates augmented reality (AR), gamification and social interactions through mobile apps to influence short- and long-term travel decisions by using a popular location-based AR gaming app, Pok mon GO, as a case study. A survey was designed and implemented to evaluate impacts of the integrated mobile apps on travel decisions using participants with different levels of familiarity and experience with Pok mon GO. Descriptive statistics and econometric model estimation results illustrate that such integrated mobile apps can be leveraged for implementing behavioral intervention strategies to influence travel decisions of subpopulations of travelers. 3 - Household Activity and Travel Patterns with Shared Autonomous Vehicles Marjan Mosslemi, University of California-Irvine, Irvine, CA, United States, R. Jayakrishnan Access to shared autonomous vehicles can create new opportunities for performing daily activities and will change traveler behavior. We explore the potential impacts by modifying the constraints of the Household Activity Pattern Problem (HAPP). The new constraints include two new specifications of the supply environment (shared mobility and driverless vehicles) and reflect following new opportunities: more vehicle resources than personal vehicles, possibility to perform activities in-vehicle, some activities that can travel. n SA30 North Bldg 221C Where To Next? Dynamic Fleet Optimization for Urban Transportation Sponsored: TSL/Urban Transportation Sponsored Session Chair: Konstantina Mellou, Massachusetts Institute of Technology, Cambridge, MA, 02141, United States 1 - Dynamically Route Deviation and Fair Cost Allocation for a Flexible Transit Service Saeid Rasulkhani, New York University, Brooklyn, NY, 11209, United States, Joseph Y. J. Chow In this study we are extending conventional MAST system which is a flexible transit system that a vehicle can deviate from it fixed path to pick up or drop off passengers. Unlike conventional MAST system, passengers may need to walk a block or so to be served rather than in their request point. Moreover, a fair cost allocation based on concept of Shapley Value from cooperative game, divides cost of deviation between served users. 2 - Routing the Boston School Buses

Sebastien Martin, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States, Dimitris Bertsimas, Arthur J. Delarue

In this talk, we focus on the major problem of school transportation, as part of a collaboration with Boston Public Schools (BPS). School transportation involves delivering students to school every morning and back home every afternoon

12