Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SC03

2 - Optimizing Water Network Mitigation Shortage and Distribution Costs Under Node Damage and Demand Uncertainty Jiangyue Gong, Texas A&M University, College Station, TX, 77840- 7140, United States, Lewis Ntaimo, Mark Alan Lawley Planning for water distribution in a network under unforeseen node damage due to natural hazards and demand uncertainty is challenging. We present a two-stage stochastic programming model for minimizing weighted water shortage and water distribution cost. The model identifies sectors to pressurize in the first stage and determines the assignment of unpressurized sectors to pressurized sectors for water delivery to satisfy demand in the second stage when the uncertainty is resolved. 3 - Modeling Wildfire Extended Attack Planning using Stochastic Programming Wildfires that are not contained after an initial response, called escaped fires, challenge decision-makers due to the high degree of temporal and spatial uncertainty surrounding fire behavior and response. We model the extended attack as a stochastic process and employ probabilistic constraints to limit response to scenarios that are feasible given resource and budgetary constraints. We present an accompanying algorithm that identifies optimal solutions while remaining computationally tractable. These solutions will inform when and how to stage and deploy resources to each fire. 4 - Data-Driven Generator Maintenance and Operations Scheduling under Endogenous Uncertainty Beste Basciftci, Georgia Institute of Technology, H. Milton Stewart School, 755 Ferst Drive NW, Atlanta, GA, 30332, United States, Shabbir Ahmed, Nagi Gebraeel In this study, our aim is to effectively model and solve the integrated condition- based maintenance and operations scheduling problem of a fleet of generators. We develop a data-driven optimization framework that explicitly considers the effect of the operations decisions on the generators` degradation levels. Since this problem involves decision-dependent uncertainties, we propose a stochastic formulation that captures the resulting endogeneity. Finally, we present computational experiments demonstrating the significant cost savings and computational benefits of the proposed approach. n SC02 North Bldg 121B Joint Session OPT/Practice Curated: Optimization under Uncertainty: Military and Cybersecurity Applications Sponsored: Optimization/Optimization Under Uncertainty Sponsored Session Chair: Rajesh Ganesan, George Mason University, Fairfax, VA, 22030, United States 1 - Critical Node Analysis and System Identification using a Discrete, General Framework for Dependency Mapping Les Servi, The MITRE Corporation, M/S M230, 202 Burlington Road, Bedford, MA, 01730-1420, United States, Erica Mason, Damon Frezza A new mission dependency mapping framework is introduced which models the relationship between an overall mission capability and its dependent component’s capability. A new genetic algorithm is presented which identifies the framework’s parameters using simulated experiments instead of the time-intensive manual alternative. A second new algorithm is presented that uses these parameters to identify the dependent components that have the greatest impact on the mission outcome. Empirical performance is reported both algorithms. 2 - Resource-enabled Pathfinding with Mandatory Waypoints and Turn Constraints Doug Altner, MITRE Corporation, 7515 Colshire Drive, M/S H617, McLean, VA, 22102, United States This presentation introduces a shortest path problem/robot motion planning problem around obstacles in a continuous space in which 1) new paths through obstacles could be created using a limited number of resources and 2) the path must also come within range of a pre-specified sequence of waypoints and satisfy turn constraints. We develop an A*-based heuristic for this problem that incorporates elements from constrained shortest path routing, theta* search, and tour routing. Computational results on over 100 test cases demonstrate our heuristic is a viable solution for this complex problem. Brittany Segundo, TAMU, 924 Sun Meadows Street, College Station, TX, 77845, United States, Lewis Ntaimo

3 - An Approximate Dynamic Programming Approach for the Financial Execution of Department of Defense Weapon System Acquisition Programs Erich D. Morman, Naval Postgraduate School, 298 Watson Street, Apartment A, Monterey, CA, 93940, United States, Rajesh Ganesan, Karla L. Hoffman Operating in a “use or lose fiscal environment, weapon system programs return millions-of-dollars each year of unspent funding. These dollars are opportunity costs to program offices representing forgone projects. The inefficiency is due to the institutional use of an inadequate myopic cash allocation policy. Using Q- learning and value function learning, we develop approximate dynamic programming (ADP) approaches to create alternative cash allocation policies. When compared to the myopic policy, our ADP models reveal that between 2% and 7% of funding is at risk of yearly “sweep-upö. The research can help program offices interested in improving overall utilization of their annual budget. 4 - Dynamic Optimization of the Level of Operational Effectiveness of a Cybersecurity Operations Center under Adverse Conditions Rajesh Ganesan, George Mason University, 4400 University Drive, Engineering Building MSN 4A6, Fairfax, VA, 22030, United States, Ankit Shah The analysts at a cybersecurity operations center (CSOC) analyze alerts generated by intrusion detection systems. There are many disruptive factors that affect the alert analysis process and as a result, adversely impact the Level of Operational Effectiveness (LOE) of the CSOC. To improve the LOE, additional resources must be called upon to assist with the alert analysis process. In a resource constrained environment, determining when and how many resources to call upon is non- trivial. In this talk, a reinforcement learning (RL) model for optimizing the LOE of a CSOC is presented. Results indicate that the RL model helps in making better decisions compared to ad-hoc practices employed at the CSOCs. n SC03 North Bldg 121C Managing Innovation Sponsored: Technology, Innovation Management & Entrepreneurship Sponsored Session Chair: Morvarid Rahmani, Georgia Institute of Technology, Atlanta, GA, 30308-1149, United States 1 - Delegated Search Impact on Startup Supply Chain Contracting and Order Allocations Berke Emre Guzelsu, Boston University, 595 Commonwealth Avenue, Boston, MA, 02215, United States, Nitin Joglekar, Pnina Feldman We examine the effects of a delegated search on startup supply chain contracting. An entrepreneurial startup partners with a collaborative supplier where the supplier needs to iterate on the product to improve the potential size of the market. At the same time, a large potential supplier may poach business from the collaborative supplier by offering a cheaper product if the order allocation it receives from the entrepreneurial startup is sufficiently large. Given these tradeoffs, we evaluate the order allocation decision an entrepreneurial firm can make to influence the collaborative supplier to invest in experimentation and examine alternative contracts to align incentives. 2 - Dynamic Innovation Contests and Information Design Sina Moghadas Khorasani, University of Utah, Salt Lake City, UT, United States, Luis Rayo, George Georgiadis We solve for the optimal design of innovation contests with multiple stages of success and an exogenous deadline. The principal selects both the award structure of the contest and its information design. 3 - Connecting Restaurants: An Exploratory Study of Customer- based Restaurant Networks Manuel Emilio Sosa, INSEAD, 1 Ayer Rajah Avenue, Singapore, 138676, Singapore, Victor Martinez-Albeniz, Clara Carrera Online customer reviews play an increasingly important role in service industries. In restaurants, links created by reviewers who visit several restaurants form a network of unobserved connections which may determine restaurants’ fate. We empirically investigate the factors that lead to the choice of visiting a restaurant, the mark given to it, and how the position of a given restaurant in the customer- based restaurant networks influences its survival. 4 - Operations in Space: Exploring a New Industry Joel Wooten, University of South Carolina, 1014 Greene St., Columbia, SC, 29208, United States, Christopher S. Tang Private, commercial spaceflight is changing the course of space exploration. We often think of innovation in terms of products, services, and (more recently) business models. The new space market presents innovation challenges in all of these areas; our paper analyzes the opportunities for novel contributions from the operations management community.

61