INFORMS 2021 Program Book

INFORMS Anaheim 2021

WB36

2 - Gpu-based Algorithms for Real-time Dial-a-ride Problems Ramesh Ramasani Pandi, Postdoc, HEC Montreal, Montréal, QC, Canada, Yossiri Adulyasak, Jean-Francois Cordeau, Louis-Martin Rousseau We study the Real-time Dial-a-ride problems (RT-DARP) and discuss how state- of-the-art GPU technology can be employed to solve RT-DARP. In this problem, the requests arrive dynamically, customers expect quick responses, and vehicles keep moving while computing assignments. Most transportation studies focus on sequential algorithms. We design a GPU-based Adaptive Large Neighborhood Search in a rolling-horizon framework for RT-DARP. The idea is to perform compute-intensive neighborhood explorations in GPU while retaining the control-intensive statements in CPU. We conduct experiments on benchmark instances from the literature and show the effectiveness of GPU on generating high-quality solutions in real-time. 3 - Dynamic Load Dependent Container Pickup and Delivery Problem with Simulation Siyuan Yao, PhD Student, University of Southern California, Los Angeles, CA, United States, Maged M. Dessouky A Multicommodity Network Flow problem is defined on a capacitated network with fixed edge costs. However, in a transportation network, edge costs depend on traffic flow in the network. Traditionally, pure mathematical formulations depict cost-flow linear relationships, which may not fit large-scale transportation networks with heterogeneous road and vehicle types. We introduce a Co- simulation-based optimization approach to estimate the network cost and provide routing decisions. 4 - Vehicle Routing Optimization with Relay: An Arc-Based Column Generation Approach Alexandria Schmid, MIT, Somerville, MA, 02143, United States, Alexandre Jacquillat, Kai Wang Several logistics providers are leveraging a new relay-based operating model: orders are routed from origin to destination through a series of pit stops and a different driver is assigned to each segment. These operations allow drivers to return home more often and offer opportunity for improved efficiency. At the same time, they raise questions on how to coordinate operations in relay networks. We propose a novel integer programming formulation to optimize the flow of trucks, drivers and orders in a time-space network. To solve it, we propose an original arc-based column generation algorithm, which generates arcs iteratively until convergence to a globally optimal solution. Results show that the algorithm outperforms traditional column generation and direct IP solutions. We conclude with practical insights from a case study on a relay logistics provider in India. WB36 CC Room 210B In Person: Behavioral Operations Contributed Session Chair: Lyudmyla Starostyuk, Metropolitan State University of Denver, Denver, CO, 80209, United States 1 - Performance Impacts of Social and Knowledge Network Alignment in Expertise Search Aaron Schecter, University of Georgia, Athens, GA, United States, Kaitlin Wowak, Ujjal Kumar Mukherjee In many organizations, complex problems are solved by effectively identifying individuals with the appropriate expertise and directing problems to them. Members of an organization are linked by two types of networks: Social networks, comprised of behavioral interactions; and knowledge networks, which represent the implicit connections between individuals’ expertise. This research examines the role of these two networks on heuristic expertise search, particularly when they converge or diverge. We study the technical service center of a large knowledge-intensive organization and identify both normative behaviors and corresponding performance impacts. 2 - Artificial Intelligence in Customer Service Operations Aykut Turkoglu, Boston University, Boston, MA, United States, Mi chelle A. Shell Companies are deploying artificial intelligence applications into service settings in a variety of ways from automating agent tasks to replacing human servers altogether. Using data from a field study, we provide early evidence that AI-based call monitoring and agent coaching improves both efficiency and customer satisfaction over human supervision alone.

3 - Do Attractive People Make a Better Deal? An Experimental Study Lyudmyla Starostyuk, PhD, Metropolitan State University of Denver, Denver, CO, United States, Yan Lang, Kay-Yut Chen The goal of our research is to shed light on the existence of an effect of seeing human faces (i.e., “face effect”) on the behavioral economic choices. We conduct a series of controlled experiments with the photographs of human faces shown in the newsvendor setting. The experimental data suggests that the human face plays the role of an environmental moderator which triggers and intensifies the social considerations such as altruism and fairness. Moreover, we find that the facial attractiveness and gender are significant motivators for the behavioral shifts in economic decisions. 4 - The Impact of Cash Transfer Participation on Child Labor in Brazil Fernanda Araujo Maciel, Assistant Professor, California State University, Sacramento, Sacramento, CA, United States The objective of this study is to assess the impact of Brazil’s Bolsa Familia conditional cash transfer program on child labor. Applying Machine Learning models to improve the estimation of the propensity score method, I analyze the effect of participating in the program on the probability of having worked in the past week and on the number of hours worked among children of 6 to 15 years old. Preliminary results show that child labor increase by 1.8 percentage points among households participating in the Bolsa Familia program. The number of hours allocated for work in this age group is not statistically different between recipients and non-recipients.

157