INFORMS 2021 Program Book

INFORMS Anaheim 2021

MB35

MB35 CC Room 210A In Person: Election Logistics/Shared Fleet General Session

MB38 CC Room 210D In Person: Innovations in Agricultural Value Chains General Session Chair: Dan Andrei Iancu, Stanford University, Stanford, CA, 94305- 7216, United States Chair: Xavier Warnes, Stanford University Graduate School of Business, Stanford, CA, 94305, United States 1 - Crop Minimum Support Price Versus Cost Subsidy: Farmer and Consumer Welfare Prashant Chintapalli, Ivey Business School, London, ON, Canada, Christopher S. Tang We analyze and compare the performance of cost subsidy and credit-based MSP. We find that (i) Although both cost subsidy and MSP induce more production, cost subsidy leads to a higher crop production than MSP; (ii) MSP improves farmer’s and consumer’s surpluses; however, cost subsidy improves consumer’s surplus but it can decrease farmer’s surplus; (iii) Although both programs achieve the same optimal net value (i.e., sum of farmer’s and consumer’s surpluses minus shortage cost and expenditure), MSP always offers higher farmer’s surplus than cost subsidy and (iv) it is beneficial to invest only in cost subsidy, in both cost subsidy and MSP, and only in MSP, when the budget availability is low, moderate, and high, respectively, so that the net surplus (i.e., sum of farmer’s and consumer’s surpluses less the shortage cost) is also maximized along with the net value. 2 - Balancing Natural Capital and Farmer Welfare: Optimal Mechanisms and Operational Implications Xavier Warnes, Stanford University Graduate School of Business, 74 Barnes Court, Apt 816, Stanford, CA, 94305, United States Many of the global agricultural commodities are produced by poor smallholders, often through illegal deforestation. The destruction of forest cover reduces the Natural Capital generated by these ecosystems. In our work we analyze these commodity supply chains and compare several supply-chain interventions that balance the Natural Capital and farmers’ welfare. MB39 CC Room 211A In Person: Editorial Positions in Journals Lessons Learned and Best Practices Panel Session Chair: Alice E. Smith, Auburn University, Auburn, AL, 36849, United States 1 - Editorial Positions in Journals Lessons Learned and Best Practices Alice E. Smith, Auburn University, Auburn, AL, 36849, United States This session will be a panel discussion about editorial roles in journals ranging from reviewing, associate editor, area/department editor, to editor-in-chief. The panelists, all experienced editors of INFORMS journals, will give their experiences and provide lessons learned and best practices. This session will have a focus on underrepresented groups and the challenges of engaging successfully with the peer reviewing and peer editing system. Important topics to be covered are how to get involved as a journal editor, how to manage your time for such a role, and how to advance your editorial career to more responsible positions. 2 - Panelist Ann Melissa Campbell, University of Iowa, Iowa City, IA, 52242- 1994, United States 3 - Panelist Archis Ghate, University of Washington, Seattle, WA, 98105, United States 4 - Panelist Katya Scheinberg, Cornell University, Ithaca, NY, 14853, United States 5 - Panelist Alejandro Toriello, ISyE Georgia Tech, Atlanta, GA, 30318, United States

Chair: Negin Shariat, University of California, Irvine, CA, United States 1 - Simulation Optimization Based Robust and Fair Allocation of Resources to Voting Locations Praveen Muthukrishnan, ISyE Georgia Tech, Atlanta, GA, United States, Benoit Montreuil, Dima Nazzal, Anjana Anandkumar, Sukanya R. Iyer, Sandro Zangiacomi The allocation of resources to voting locations influences throughput capacity and waiting time distribution across locations in a political territory (e.g. county), where each location is targeted to serve a subset of the territory (e.g. precinct). Usually subject to tight budget constraints and having significant impact on multi- criteria performance, election boards allocate poll pads, ballot marking devices and scanners using simple ratios such as voters-per-resource. These do not account for local differences in voter turnout, hour-of-the-day voter arrival, and poll time distributions. We introduce a simulation-optimization approach that maximizes robust wait-time performance, enforces fairness across voting locations, and respects budget constraints. We benchmark our approach against actual allocation in Fulton County for the 2020 US Election. 2 - Hypothetical Networks for Analysis of Transit and Shared Mobility Systems Negin Shariat, University of California, Irvine, CA, United States, R. Jayakrishnan Transportation planners usually consider many geographical or social demographic factors when designing frameworks for transit system or a shared mobility option for an area. These frameworks are limited in application and constrained to the situational context in the study area. Therefore, contextual transferability of studies is often questionable. We present an unlimited hypothetical but realistic network database that includes the network and its supply and demand to help planners in testing their algorithms and taking different aspects of the area into consideration. Finally, we generate a set of scalable networks with desired assumptions and topologies and test transit planning approaches as well as a shared mobility algorithm to evaluate their capabilities for application in different networks. MB36 CC Room 210B In Person: Demand Management for Last-mile Logistics and On-demand Mobility General Session Chair: Vienna Klein, Neubiberg 1 - Policy-based Dynamic Pricing in Shared Mobility Systems Matthias Soppert, Bundeswehr University-Munich, Munich, Germany, Claudius Steinhardt Shared mobility systems have become a wide-spread alternative within the inner- city mobility. Modern systems offer one-way trips, which yield high flexibility to the customer but also cause imbalances between supply and demand that need to be rebalanced for profitable operation. Pricing has turned out to be a promising means. We consider the on-line problem of a shared mobility system provider to simultaneously set discrete minute prices for all zones of the operating area. The action space of this stochastic dynamic decision problem grows exponentially with the number of zones, such that value-based approaches do not scale. Instead, we propose a policy-based approach, adapted from the realm of deep reinforcement learning, which can handle the large actions space. Preliminary results indicate that our approach surpasses the optimal static as well as dynamic benchmarks. 2 - Implications of Different Dynamic Modelling Approaches for Integrated Demand Management and Vehicle Routing Problems Vienna Klein, Bundeswehr University Munich, Neubiberg, Germany, David Fleckenstein, Claudius Steinhardt, Robert Klein Demand control problems in the field of vehicle routing are characterized by a stream of customers arriving dynamically over a booking horizon and requesting logistical services which are fulfilled by a given fleet of vehicles. Demand management methods can be applied to exploit heterogeneous customer preferences in order to optimize the booking process with the aim of maximizing total profit. As the quality of demand management decisions depends to a large extent on an accurate estimation of opportunity cost, we formalize its definition specifically for vehicle routing applications. Furthermore, we discuss their properties for different dynamic modelling approaches to derive and discuss implications for approximate dynamic programming solution approaches.

54