2015 Informs Annual Meeting

WD70

INFORMS Philadelphia – 2015

2 - Analytical Stochastic Link Transmission Model Suitable for Large-scale Analysis Jing Lu, Massachusetts Institute of Technology, Cambridge, MA, United States of Americalucifinil.lj@gmail.com, Carolina Osorio We formulate a model that builds upon the previously formulated analytical stochastic network loading model of Osorio and Flotterod (2013). The proposed model has a complexity that is linear, rather than cubic, in the space capacity of the links in the network. It is a scalable approach suitable for large-scale network analysis. The model is validated versus stochastic simulation results, and is then used to analyze a network with intricate network dynamics. 3 - A Graph-based Approach to Measure the Efficiency of Urban Taxi Service System Xianyuan Zhan, PhD Candidate, Purdue University, 550 Stadium Mall Drive, School of Civil Engineering, West Lafayette, IN, 47906, United States of America, zhanxianyuan@purdue.edu, Xinwu Qian, Satish V. Ukkusuri This study investigates the efficiency level of the taxi service system using real world large-scale taxi trip data from New York City. Two large-scale optimization problems are formulated and transformed into equivalent graph problems to find the theoretical optimal strategy that minimizes the cost of vacant trips, and results in minimum number of taxis required to satisfy all observed trips. Huge performance gap is observed between current system and the optimal system, which suggests the potential gain by sharing system-wide information among taxi drivers and passengers. 4 - Combining Data-driven and Model-driven Approaches for Traffic Simulator Calibration Problems Kevin Zhang ,Massachusetts Institute of Technology, Cambridge MA, United States of America, kzhang81@mit.edu In this presentation, we approach the problem of parameter calibration of stochastic traffic simulators through the use of Kalman filtering. We build upon an established Kalman filtering approach by incorporating network-specific structural information supplied by an analytical queueing model. The approach is benchmarked versus other traditional calibration methods. Results on low- dimensional calibration problems are presented. Real-Time Traffic Monitoring and Control Sponsor: TSL/Intelligent Transportation Systems (ITS) Sponsored Session Chair: Xuan Di, University of Michigan, Ann Arbor, MI, United State of America, sharondi@umich.edu 1 - Psychological Benefits of Real-Time Travel Information on Route Choice Behavior – Analysis of Interactive Driving Simulator Experiment Data Dong Yoon Song, Purdue University, West Lafayette, IN, United States of America, song50@purdue.edu, Srinivas Peeta This study proposes a comprehensive framework to understand the psychological mechanisms associated with real-time travel information provided. A structural equation model with latent variables is presented to address the driver’s psychological processes associated with real-time information based on revealed preference data from interactive driving simulator experiments and the associated survey data. Insights from the interactive driving simulator experiments will also be discussed. 2 - Determining Optimal Toll Gantry Locations on Tollway Ruoyu Liu, CDM Smith, 8140 Walnut Hill Ln # 1000, Dallas, TX, United States of America, happylry@gmail.com, Worapong Hirunyanitiwattana Toll Gantry Locations can affect traffic flow and toll revenue on tollway. The model maximizes toll revenue and also minimizes the travel cost. Genetic algorithm is used to search the optimal solution based on result from a traffic assignment model. 3 - Real-Time Lane-Based Traffic State Estimation and Prediction Kerem Demirtas, Arizona State University, Tucson, AZ, United States of America, kerem.demirtas@asu.edu, Pitu Mirchandani, Xuesong Zhou Given new lane-based ITS applications such as speed controls and managed lanes, one needs to estimate and predict the state of the traffic in each lane of a highway. Based on Newell’s 1993 Simplified Kinematic Wave Model, and the p- detector interpretation of Daganzo, we have an approach to estimate past trajectories from Lagrangian flow-density measurements and subsequently predict short-term trajectories using simplified behavior models. Analysis using real and simulated data is discussed. WD69 69-Room 201C, CC

4 - Desirability Measures and Discovery Analysis for Multi-Modal Services Based on Daily Trajectory Pattern Data Yi-Chang Chiu, University of Arizona, Tuscon, AZ, United States of America, chiu@email.arizona.edu, Ali Arian This talk focuses on presenting a computational method to estimate the desirability measures for various transportation modes available to a traveler based on known multi-day daily GPS trajectory pattern data. Algorithmic details and case studies are presented. WD70 70-Room 202A, CC Vehicle Routing III Contributed Session Chair: Alexander Düge, TU München, Arcisstr. 21, München, Germany, alexander.doege@tum.de 1 - Appointment Scheduling with Multiple Providers and Stochastic Service Times Mohamad Soltani, University of Alberta, PhD Office, Business Building, Edmonton, AB, T6G 2R3, Canada, soltani@ualberta.ca, Michele Samorani, Bora Kolfal We consider a multi-server appointment scheduling problem in which patients may not show up, and those who show up require stochastic service times. We model this problem to evaluate each candidate solution. We statistically find some properties for the optimal or near optimal solutions, and design a simulation optimization approach using these properties. We also propose a heuristic algorithm, and validate its performance by comparing its result with our simulation optimization approach. 2 - Exact Optimization Frameworks for Time-consistent Routing Anirudh Subramanyam, Graduate Student, Carnegie Mellon University, DH3122, 5000 Forbes Ave, Pittsburgh, PA, 15213, United States of America, asubramanyam@cmu.edu, Chrysanthos Gounaris We present exact approaches for the Consistent Traveling Salesman Problem, wherein arrival-time consistency across multiple periods is enforced for each customer. Our first approach is based on branch-and-cut. We compare alternative formulations and propose new valid inequalities. Our second approach is based on decomposition into single-period TSPTWs, where consistency is enforced by tightening the windows during the search. We compare our approaches on benchmark instances derived from TSPLIB. 3 - Rollout Algorithm for The Dynamic Vehicle Routing Problem In City Logistics Gitae Kim, Assistant Professor, Hanbat National University, N4 - 207, 125 Dongseo-daero, Yuseong-gu, Daejeon, 305-719, Korea, Republic of, gitaekimemail@gmail.com, Yew Soon Ong, Taesu Cheong This paper suggests a dynamic vehicle routing problem (DVRP) model in city logistics. Depending on the traffic conditions, the travel time between two nodes, particularly in a city, changes both dynamically and stochastically over time. In this paper, we propose a Markov decision process (MDP) model for the DVRP with the travel time dynamics under traffic congestion in the city area and adopt the rollout algorithm to solve the problem. 4 - Vehicle Routing with Flexible Delivery Locations Alexander Düge, TU München, Arcisstr. 21, München, Germany, alexander.doege@tum.de, Daniel Gartner, Markus M. Frey We present a new extension of the vehicle routing problem (VRP): The VRP with flexible delivery locations and time points (VRPFLTP). In the VRPFLTP, a customer not only corresponds to exactly one location but has to be served at one out of a set of possible capacitated locations. We develop an adaptive large neighborhood search which demonstrates high quality solutions within short computation times. In†our economic analysis, we trade off location flexibility on a†variety of performance metrics.

475