Informs Annual Meeting 2017

TA66

INFORMS Houston – 2017

TA67

3 - Approximate Dynamic Programming for the Multi-Period Technician Scheduling with Experience-based Service Times and Stochastic Customers (MTSP-ESTSC) Xi Chen, Beijing Foreign Studies University, Beijing, China, chenxi0109@bfsu.edu.cn, Barrett Thomas, Mike Hewitt In this work, we explore the value of integrating future information into the current period decision-making process for the MTSP-ESTSC. We propose an approxi- mate dynamic programming solution approach that uses a basis function to reduce the dimensionality of the state. We benchmark the performance of the ADP solution ap- proach with the basis function against a one-day lookahead approach that minimizes the recent two periods’ costs. Our computational results demonstrate the value of proposed approach. 4 - A Benders Decomposition Method for Two-stage Stochastic Network Design Problems Walter Rei, University of Quebec-Montreal, 10673 Larose, Montreal, QC, H2B 3C4, Canada, rei.walter@uqam.ca, Ragheb Rahmaniani, Teodor Gabriel Crainic, Michel Gendreau In this presentation, we describe a Benders decomposition algorithm capable of efficiently solving large-scale instances of the multi-commodity capacitated network design problem with demand uncertainty. This problem is relevant to a wide gamut of real-world applications in the telecommunications, transportation and logistics fields. To obtain a state of the art Benders method, we propose various acceleration techniques, including the use of cutting planes, partial decomposition, heuristics, stronger cuts, reduction and warm-start strategies. Extensive computational experiments conducted on benchmark instances confirm the superiority of the proposed algorithm over existing ones. 371A Network Design in Freight Logistics Sponsored: Transportation Science & Logistics Sponsored Session Chair: Lokesh Kalahasthi, Rensselaer Polytechnic Institute, Troy, NY, United States, kalahl@rpi.edu 1 - Scheduling Optimization of Local Farmers Sefakor Fianu, North Carolina A&T.State University, 1601 .Market St, Greensboro, NC, 27401, United States, sfianu@aggies.ncat.edu, Lauren Berrings Davis Local farmers face challenges such as; seasonality, transportation, food safety, and packaging requirements that prevent them from meeting the demand of institutions like universities. This research is part of the University Food Systems project aimed at connecting local farmers and universities in North Carolina. We solve the transportation problem by developing weekly schedules that minimizes the distance traveled by farmers to drop off their produce at crossdocking facilities. The schedules addresses capacity, collection frequency and food spoilage constraints. 2 - The Dynamic Shortest Covering Path Problem Jimena Pascual, Pontificia Universidad Catolica de Valparaiso, Av Brasil 2241, Casilla 4059, Valparaiso, 2362807, Chile, jimena.pascual@pucv.cl, Andrea Leticia Arias, Dario Canut De Bon, Ricardo Gatica, Timothy I. Matis We consider a variant of the Shortest Covering Path Problem (SCPP) in which the arc costs change over time due to dynamic environmental conditions. In this problem, the aim is to determine a minimum cost path from a starting to an ending position, such that all the nodes in the network are covered by at least one node selected to be in the path. We present an integer programming formulation for the Dynamic SCPP and explore solution methodologies. 3 - Freight Demand Synthesis with Mode Choice; A Combined Estimation Procedure Lokesh Kumar Kalahasthi, Rensselaer Polytechnic Institute, 22, College Ave,, Troy, NY, 12180, United States, kalahl@rpi.edu, Jose Holguin-Veras, John E.Mitchell This research develops a combined model for freight demand synthesis (FDS) that incorporates the estimation of modal split between rail and truck. A gravity model is adopted for the estimation of trip distribution; a binary logit model for modal split between rail and truck; and Noortman and van E’s model is used for empty trips. It is assumed that the total productions and attractions, the network data, and the link flows by both modes are available. Application on a sample network shows that, the model provides reasonably good estimates for O-D table and modal split. This research serves as a potential tool for transportation planners in evaluating various policy outcomes. TA66

371B High-dimension Spatio-Temporal Analysis II Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Shyam Ranganathan, Virginia Tech, shyam81@vt.edu Co-Chair: Jian Liu, University of Arizona, Tucson, AZ, 85721, United States, jianliu@email.arizona.edu 1 - A Class of Bayesian Hierarchical Models with Conjugate Full-conditional Distributions for Dependent Data Jonathan Bradley, Florida State University, Tallahassee, FL, United States, bradley@stat.fsu.edu We introduce a class of models for Bayesian analysis of high-dimensional dependent data. In particular, we introduce something we call the “conjugate multivariate distribution.” Furthermore, we provide substantial theoretical and methodological development including: results regarding conditional distributions, an asymptotic relationship with the multivariate normal distribution, conjugate prior distributions, and full-conditional distributions for a Gibbs sampler. The results in this manuscript are extremely general, and can be adapted to many different settings. We demonstrate the methodology through several illustrations. 2 - Estimation of Sparse Vector Autoregressive Moving Averages Ines Wilms, KU. Leuven, Naamsestraat 69, Leuven, 3000, Belgium, Ines.Wilms@kuleuven.be, Sumanta Basu, Jacob Bien, David Matteson The Vector Autoregressive Moving Average (VARMA) model is a fundamental tool for modeling multivariate time series. However, as the number of time series increases, the VARMA model becomes heavily overparameterized. For such high- dimensional VARMA models, estimation is generally considered intractable, leading people to favor simpler approaches such as Vector Autoregressive (VAR) models. We propose a new framework for learning parsimonious VARMA models in high-dimensional settings and show that it has good forecast accuracy both in simulation and in various application domains. 3 - A Statistical Modeling Approach for Large-scale Spatio-temporal Data Xiao Liu, Assistant Professor, University of Arkansas, Department of Industrial Engineering, 4207 Bell Engineering Center, 1 University of Arkansas, Fayetteville, AR, 72701, United States, liuxiao314923@gmail.com The research presents a statistical modeling approach for large-scale spatio- temporal data. In particular, we show the connection of the statistical model to physical models which are given by Stochastic Partial Differential Equations. 371C Manufacturing Data Analytics Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Hongyue Sun, University at Buffalo, Buffalo, NY, 1, United States, hongyues@buffalo.edu 1 - Heterogeneous Recurrence Monitoring of Dynamic Transients in Ultraprecision Machining Processes Chen Kan, Pennsylvania State University, University Park, PA, United States, cjk5654@psu.edu, Changqing Cheng, Hui Yang In-situ monitoring and control of process variations are important for quality assurance in ultraprecision machining (UPM) processes. However, conventional approaches are limited in their ability to address the complex dynamics in the nonlinear and nonstationary processes. This paper presents a heterogeneous recurrence monitoring approach to detect dynamic transients in UPM processes. Experimental results showed that the proposed approach effectively detects transitions with a small magnitude in the Lorenz system, and identifies the shift from stable cutting to unstable cutting in the real-world study of UPM processes. TA68

294