INFORMS 2021 Program Book

INFORMS Anaheim 2021

WB08

2 - Sequential Search with Bidding Cagin Uru, Duke University, Durham, NC, United States, David Brown We consider a sequential search problem in which a decision maker sequentially explores and bids on a set of disappearing alternatives with a priori unknown values to obtain the best alternative. This problem can be formulated as a stochastic dynamic program, but it is difficult to solve as the state space is both high-dimensional and continuous. In this paper, we study a relaxation with an infinite number of alternatives and obtain a threshold policy that does not recall any previously explored alternative. Analyzing its performance in the original problem, we not only show that the threshold policy is asymptotically optimal for problems with many alternatives but also characterize the convergence rate. For problems where calculating the threshold is challenging, we propose an approximation scheme based on the discretization of the state space and prove its convergence. 3 - A Real Options Approach to Risk Management for a Small Business Innovation Research Portfolio Jeremy Eckhause, RAND Corporation, Arlington, VA, 22202-5005, United States, Andrea Belz, Fernando Zapatero, Richard Terrile An important class of R&D investments in public contexts is one in which investments are staggered in multiple stages, with relatively modest early funding for the selected proposals and higher levels subsequently made available for a subset. We present a method based on a real options approach to select a portfolio of proposals in each stage, applied to the NASA’s Small Business Innovation Research (SBIR) portfolio. Among the practical implications of our analysis, the model indicates that a strong increase in the funding of the smallest firms has only minimal impact on the overall portfolio value. TE42 CC Room 212B In Person: Disaster and Disruption Management Contributed Session Chair: Jorge Huertas, Georgia Institute of Technology, Atlanta, GA, 30309, United States 1 - Post-Hurricane Damaged Timber Management Problem Using Bilevel Model Formulation Amin Aghalari, Mississippi State University, Starkville, MS, United States, Mohammad Marufuzzaman, Badr Aladwan, Shaun Tanger, Bruno Silva This study proposes a bi-level mixed-integer linear programming model to optimize different critical decisions (e.g., purchasing, storage, and transportation decisions) of a post-hurricane damaged timber management problem. Further, this study develops two exact solution methods, namely, the enhanced Benders decomposition and the Benders-based branch-and-cut algorithms to efficiently solve the model in a reasonable timeframe. We use 15 coastal counties in southeast Mississippi to visualize and validate the algorithms’ performance. 2 - Optimal Selection of Pre-event Short and Long-term Mitigation Strategies for Flooding Hazards We propose a mathematical model to study the effects and tradeoffs associated with pre-event short-term and long-term mitigation strategies to minimize the economic loss associated with flooding hazards. We illustrate the capabilities of the model with a case study on Lumberton, NC. Lumberton has been affected by severe flooding events with significant recurring economic loss. The model uses the cost from a portfolio of mitigation strategies, each representative of a different mitigation strategy, and the resulting flood-induced monetary losses corresponding to each strategy. Finally, the optimal flood mitigation plan for buildings is provided based on a mitigation budget constrained. 3 - Use of Ships for Fuel Emergency Distribution on Islands Vahid Eghbal Akhlaghi, University of Iowa, Iowa City, IA, United States, Ann Melissa Campbell We present a mixed-integer programming model to examine the strategies for using ships to supplement the fuel supply on islands after a major disaster. The problem is motivated by practices proposed by FEMA after recent hurricanes in the Caribbean. The model presented includes decisions about routing ships to ports and assignment of fuel dispensing sites to ports to minimize the latest time a fuel distribution site receives its required fuel supply. An extension of the model to consider the use of standby ships is introduced. After proving the NP-hardness of the problem, we derive structural properties, lower bounds, and valid inequalities. A case study based on real data for Puerto Rico is presented. Himadri Sen Gupta, University of Oklahoma, Norman, OK, United States, Omar Magdy Nofal, Andres David Gonzalez, Charles D. Nicholson, John W. van de Lindt

4 - Modeling of Covid-19 Trade Measures on Essential Products: A Multiproduct, Multicountry Spatial Price Equilibrium Framework

Mojtaba Salarpour, University of Massachusetts-Amherst, Amherst, MA, United States, AnnaB. Nagurney, June Dong

We develop a unified variational inequality framework in the context of spatial price network equilibrium problems that handles multiple products with multiple demand and supply markets in multiple countries as well as multiple transportation routes. The model incorporates a plethora of distinct trade measures, which is particularly important in the pandemic, as PPEs and other essential products are in high demand, but short in supply globally. In the model, product flows as well as prices at the supply markets and the demand markets in different countries are variables that allows us to seamlessly introduce various trade measures, including tariffs, quotas, as well as price floors and ceilings. 5 - Multi-objective Community Resilience Optimization with CGE Modeling for Memphis Metropolitan Statistical Area Rafia Bushra, University of Oklahoma, Norman, OK, United States Natural hazards have the potential to cause billions of dollars of damage, create major disruptions in key elements of communities worldwide, and drive complex outcomes such as population dislocation, unemployment rates, threats to household income, etc. In this presentation, we consider the Memphis Metropolitan Statistical Area lying within the New Madrid seismic zone. We implement a community resilience multi-objective optimization model that leverages a reversed engineered computable general equilibrium model derived information to capture system-wide impacts to enhance decision-making. 6 - Large-scale Zone-based Evacuation Planning: Generating Convergent and Non-preemptive Evacuation Plans via Column Generation Jorge A. Huertas, Georgia Institute of Technology, Atlanta, GA, United States, Pascal Van Hentenryck In zone-based evacuations, the evacuated region is divided into zones, and vehicles follow the single evacuation path assigned to their corresponding zone. Ideally, these evacuation paths converge at intersections to reduce driver hesitation; and non-preemptive schedules ensure that the evacuation of a zone, once it starts, proceeds without interruptions. We present a macroscopic optimization model to produce convergent and non-preemptive evacuation plans. Furthermore, we decompose our model and use a column-generation algorithm to solve it in real large-scale evacuation scenarios. Finally, we use a microscopic traffic simulator to evaluate the quality of the generated plans.” In Person: APS Fairness in Sequential-Decision Making General Session Chair: Siddhartha Banerjee, Cornell University, Ithaca, NY, 14853- 3801, United States Co-Chair: Sean Sinclair, Cornell University, Ithaca, NY, 14853, United States 1 - A Markovian Arrival Stream Approach To Stochastic Gene Expression In Cells Brian Fralix, Clemson University, Clemson, SC, 29642-8005, United States, Mark Holmes, Andreas Lopker We analyze an abstraction of the stochastic gene expression model studied recently in Fromion et al. (SIAM Journal of Applied Math, 2013) and Robert (Probability Surveys, 2019) using techniques from the theory of point processes, and the theory of matrix-analytic methods. In the model we consider, both the activity of a gene and the creation of mRNA are modeled with an arbitrary Markovian arrival process. This modification is important, as Markovian arrival processes can be used to approximate many types of point processes on the nonnegative real line. Wednesday, 7:45AM-9:15AM WB08 CC Room 303C

149