INFORMS 2021 Program Book

INFORMS Anaheim 2021

SC11

SC12 CC Room 304D In Person: AAS Best Student Presentation Competition Award Session Chair: Kai Wang, MIT Sloan School of Management, Boston, MA, 02215-4212, United States 1 - Trajectory Planning for Mission Survivability of Autonomous Vehicles in Moderately to Extremely Uncertain Environments Fanruiqi Zeng, Georgia Institution of Technology, Atlanta, GA, United States, Husni R. Idris, John-Paul Clarke In this work, we propose a receding horizon control strategy with novel trajectory planning policies that enable dynamic updating of the planned trajectories of autonomous vehicles operating in environments where potential conflicts are, from a statistical perspective, either partially known or completely unknown. The proposed policies utilize two metrics: (1) the number of feasible trajectories; and (2) the robustness of the feasible trajectories. We measure the effectiveness of the suggested policies in terms of mission survivability. Our findings have significant implications for achieving safe aviation autonomy. 2 - Traffic Management and Resource Allocation for UAV-Based Parcel Delivery in Low Altitude Urban Airspace Ang Li, University of California-Berkeley, Berkeley, CA, 94720-2392, United States As the development in e-commerce presents the major driver for drone-based deliveries, the need for and importance of efficiently managing UAV traffic in urban airspace is arising. This research proposes a framework of UAV system traffic management in the context of parcel delivery in low-altitude urban airspace, including clustering-based UAV path planning, systematic UAS traffic management with conflict resolution, and mechanism design for airspace resource allocation. Extensive numerical analysis is conducted with San Francisco as the case study area. Our results show the effectiveness of the proposed framework and the scalability of traffic management model. 3 - Using Submodularity Within Column Generation to Solve the Flight-to-Gate Assignment Problem In this paper, we provide a column generation-based approach for solving the airport flight-to-gate assignment problem, where the goal is to minimize the on- ground portion of arrival delays. Specifically, we use a set covering formulation for the master problem and decompose the pricing problem such that each gate is the basis for an independent pricing problem. We use a combination of an approximation algorithm based on the submodularity of the underlying set and dynamic programming algorithms to solve the independent pricing problems. We also design and employ a rolling horizon method and block decomposition algorithm to solve large-sized instances. 4 - Data-Driven Robust Aircraft Assignment to Minimize Delay Propagation Wei Liu, University of North Carolina at Chapel Hill, Chapel Hill, NC, 27516-8361, United States, Vinayak V. Deshpande, Vidyadhar Kulkarni We propose a new approach to reduce the delay propagation by optimizing the assignment between incoming and outgoing flights flown by an airline at a given airport. Specifically, we provide a data-driven approach to estimate the arrival delay distribution, and then derive several assignment policies based on the estimated distribution. We show that the assignments derived from the data- driven approach can offer a verifiable improvement compared to the optimal assignment (FIFO) derived in the deterministic setting by using the real data of Delta Airlines at Atlanta airport. 5 - Sequential Prediction of Flight Anomaly Using Real-time Data: A Case Study for Go-around Lu Dai, University of California, Berkeley, Berkeley, CA, United States Lu Dai, National Center of Excellence for Aviation Operations Research, Berkeley, CA, United States, Mark M. Hansen Disruptions caused by flight anomaly increase the workload for operators, and leading to more developing risks. As a case study, we encapsulate predictive analytics to provide real-time sequential prediction of go-arounds by fusing multiple real-time data sources and developing learning models to estimate the probability of go-arounds near the airport. We demonstrate our framework on a real-time feed emulator and compare the performance of learning models trained on datasets with different proportions of synthetic go-around sequences which are generated by different augmenting techniques. This research accelerates predictive analytics for aviation safety in the real-time arena. Yijiang Li, Georgia Institute of Technology, Atlanta, GA, United States, John-Paul Clarke, Santanu Subhas Dey

SC11 CC Room 304C In Person: Fairness in Operations General Session Chair: Vahideh Manshadi, Yale University, Quincy, MA, 02169-4688, United States Co-Chair: Rad Niazadeh, Chicago Booth School of Business, CA, 94305-5008, United States Co-Chair: Scott Rodilitz, Yale, New Haven, CT, 06511-2572, United States 1 - Bias and Discrimination in Machine Learning: Automated Employment Screening as a Case Study Manish Raghavan, Cornell University The use of algorithmic decision-making in socially consequential domains has raised fundamental questions over what constitutes “fair” or “unbiased” decision- making. In this talk, I will situate these questions in the context of data-driven hiring decisions. I’ll give an overview of how algorithms are used in hiring, discuss how vendors of algorithmic tools operationalize notions of non- discrimination, and map the legal and policy challenges in that arise in combatting algorithmic discrimination. 2 - Fairness in Hiring and Beyond Swati Gupta, Fouts Family Early Career Professor, Georgia Institute of Technology, Atlanta, GA, 30332, United States, Jad Salem, Deven R. Desai The introduction of automation into the hiring process has put a spotlight on a persistent problem: discrimination in hiring on the basis of protected-class status. Left unchecked, algorithmic applicant-screening can exacerbate pre-existing societal inequalities and even introduce new sources of bias; if designed with bias- mitigation in mind, however, certain group-aware interventions can be construed as illegal due to requirements of U.S. anti-discrimination law. In this work, we will focus on selection algorithms used in the hiring process (e.g., resume-filtering algorithms) given access to a “biased evaluation metric”, i.e., we assume that the method for numerically scoring applications is inaccurate in a way that adversely impacts certain demographic groups. We will conclude the talk by ways to argue legal feasibility of our proposed interventions. 3 - Discrimination, Diversity, and Information in Selection Problems Faidra Monachou, Stanford University, Stanford, CA, 94305-7224, United States We study the role of information, access, and privilege in capacity-constrained selection problems with fairness concerns. We introduce a theoretical framework that formalizes the trade-off between the informativeness of a feature and its exclusionary nature when members of different social groups have unequal access to this feature.We extend the model to study the role that differential privilege and the correlation between skill and privilege play in discrimination.Our framework finds a natural application to recent policy debates on dropping standardized testing in admissions, soft affirmative action, and randomized admission policies. 4 - Order Symmetry of Assignment Mechanisms Rupert Freeman, University of Virginia, Charlottesville, VA, 22901, United States, Geoffrey Pritchard, Mark Curtis Wilson We introduce a new average-case fairness criterion, order symmetry, for assignment mechanisms. We argue for its importance, clarify its relationship to other axiomatic properties, and analyze the performance of the Top Trading Cycles (TTC), Serial Dictatorship, Naive Boston and Adaptive Boston mechanisms with respect to order symmetry. We define some basic measures of order bias, or lack of order symmetry. Low order bias is a necessary but not sufficient condition for high egalitarian welfare. We study order bias under several preference distributions and show that for sufficiently symmetric distributions, TTC is order symmetric while the other three mechanisms we consider are not.

18