Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TE36

3 - Airline Passenger Route Share Forecast Xufang Zheng, Iowa State University, 537 Bissell Road #2362, Ames, IA, 50010, United States, Chia-Mei Liu, Peng Wei Airline passenger route share(directShare) is the ratio of direct passengers to total passengers on O&D level. It is an important feature of passenger flow distribution. DirectShare is an O&D specific feature, which is highly correlated with quarterly lag. Various supervised learning methods are carefully explored. The best model is gradient boosting machine(GBM), which has better prediction performance than FAA TAF-M directShare forecast model. Category based learning is newly proposed, which provides better prediction performance than GBM. The C- basedapTC model is the best category based learning model, which can provide a long term directShare forecast with less fluctuations. 4 - Game Theoretic Modeling for Airline Frequency Competition in Large Networks Reed Harder, Dartmouth, 14 Engineering Drive, Hanover, NH, 03755, United States, Vikrant Vaze Airlines set daily flight frequencies on origin-destination segments across their networks. These decisions are made in competition with other airlines and in coordination with potential connecting flights on other segments, and have significant implications for the efficient functioning of the air transportation system. Game theoretic models have been used to analyze incentives in airline resource allocation, but often suffer from limited computational tractability and predictive accuracy in large networks with many connections. We present approaches for tractable game theoretic modeling in these large networks using approximate methods. 5 - Strategic Behaviors in Airport Capacity Allocation Mechanisms Weilong Wang, Purdue University, West Lafayette, IN, United States, Alexandre Jacquillat, Vikrant Vaze We develop an original bi-level game-theoretic approach to identify opportunities for strategic behaviors by non-atomistic users (e.g. airlines) in non-monetary mechanisms for infrastructure (e.g. airport) capacity allocation. We show that gaming opportunities are limited under a primary mechanism, where capacity is allocated to individual flights. In contrast, airlines may have strong gaming opportunities in a secondary mechanism, where capacity is allocated to airlines who may then swap their own flights. We present computational results comparing the overall performance of both mechanisms. n TE36 North Bldg 224B Route Optimization for Drones Emerging Topic: Robotics, Drones and Autonomous Vehicles in Logistics Emerging Topic Session Chair: Andrea Leticia Arias, Texas Tech University, Lubbock, TX, 79424, United States 1 - A Heuristic Approach to Path-planning for Delivery Drones in Anisotropic Medium Abhishake Kundu, Texas Tech University, 201 Indiana Avenue, # D210, Lubbock, TX, 79415, United States, Timothy Matis The purpose of this research is to consider an efficient heuristic to navigate parcel delivery drones (in conjunction with trucks) in direction dependent uniform wind-fields. Constraints on specific relative velocities and available battery power for the drones makes this an excellent utilitarian extension on developing literature. 2 - Route Optimization of Unmanned Area Vehicle with Radio Frequency Identification Interrogator Victoria Carson, California State Polytechnic University, Dept of Industrial & Manufacturing Eng, San Luis Obispo, CA, 93407, United States, Tali Freed, Neil Wolfe, Jonathon Scott The Close Enough Traveling Salesman model is used to optimize the route of an unmanned aerial vehicle (UAV) equipped with a radio frequency identification (RFID) interrogator. The UAV’s mission is to identify assets located in a given area within the flight time of a single battery charge. Reported use cases include cattle in grazing pastures and oil drilling equipment. 3 - A Deterministic Two-level Integrated Inventory Approach for We consider the problem of optimizing the management of batteries at a drone battery swap station which allows for instantaneous swapping of depleted batteries for fully-charged ones. Since drones are powered by batteries with limited flight range, the decisions made with regards to battery charging and replacement actions determine which geographically diverse locations are able to be serviced and when. We use a deterministic two-level integrated inventory model where the first and second inventory levels are battery charge and capacity, respectively. We develop a heuristic that minimizes average weighted delivery time and perform tests to deduce policy insights. Inventory Management at a Drone Battery Swap Station Olivier Kwizera, University of Arkansas, Fayetteville, AR, United States, Sarah G. Nurre

n TE37 North Bldg 225A

Joint Session APS/Opt-Uncert: Interfaces between Applied Probability and Robust Dynamic Optimization Sponsored: Applied Probability Sponsored Session Chair: Chaithanya Bandi, 1987, Evanston, IL, 60208, United States 1 - Optimal Approximations for Two-stage Adjustable Robust Optimization Under Budgeted Uncertainty Omar El Housni, Columbia University, New York, NY, 10027, United States, Vineet Goyal We study the performance of affine policies for two-stage adjustable robust optimization problem under a budget of uncertainty set. The two-stage adjustable robust optimization problem is hard to approximate within a factor better than $\Omega( \frac{\log n}{\log \log n})$ for budget of uncertainty sets where $n$ is the number of decision variables. We show that surprisingly affine policies provide the optimal approximation for this class of uncertainty sets that matches the hardness of approximation; thereby, further confirming the power of affine policies. We also provide a faster algorithm to compute near-optimal affine policies. 2 - Robust Queueing Approach to Optimal Control of Fork-join and Replication Systems Chaithanya Bandi, Kellogg School of Management, Northwestern University, 2211 Campus Dr, room 4169, Evanston, IL, 60208, United States We consider the problem of control and analysis of fork-join queues. 3 - A Robust Queueing Network Analyzer Based on Indices of Dispersion Wei You, Columbia University, New York, NY, 10032, United States, Ward Whitt We present a robust queueing network analyzer (RQNA) algorithm to approximate the steady-state performance of a single-class open queueing network of single-server queues with Markovian routing, allowing general interarrival and service distributions. The RQNA algorithm includes subroutines to (1) approximate system performance measures using the index of dispersion for counts (IDC) of the arrival process at each station, (2) calculate or estimate the IDC’s for external flows, (3) solve systems of linear equations to approximate the IDC’s for internal flows and (4) eliminate customer feedback. Effectiveness of the RQNA algorithm is supported by heavy-traffic limits and simulations. 4 - Robust Maximum Likelihood Estimation Omid Nohadani, Northwestern University, 2145 Sheridan Road, Technological Institute M233, Evanston, IL, 60208-3119, United States, Dimitris Bertsimas In many applications, statistical estimators serve to derive conclusions from data. However, the conclusions are dependent on uncertainties in the data. We use robust optimization principles to provide robust maximum likelihood estimators that are protected against data errors. We provide efficient local and global search algorithms to compute the robust estimators. The performance is demonstrated on simulated data and on a large set of clinical radiation therapy data, where robust estimators offer more reliable decisions. This approach is general and applicable to a broad range of problems. n TE38 North Bldg 225B Joint Session APS/Practice Curated: Discrete Convexity and its Application Sponsored: Applied Probability Sponsored Session Chair: Linwei Xin, University of Chicago, Chicago, IL, 60637, United States Co-Chair: Xin Chen, UIUC, Urbana, IL, 61801, United States 1 - Optimal Inventory Management of a Blood Center Yixuan Liu, University of Texas, 2110 Speedway, B6000, IROM, Austin, TX, 78705, United States, Youyi Feng, Guoming Lai We study inventory management for a blood center that periodically collects whole blood and processes it into multiple blood products including red blood cell and platelet. With limited lifetimes, blood products are stored to satisfy random demands or sold in clearance sales before becoming outdated. The blood center also has an option to replenish red blood cells from an alternative source. Applying the concept of multimodularity, we characterize the structural properties of the value function and the optimal strategies. We demonstrate that the replenishment and salvage decisions can be coordinated to optimize the inventory system and discuss the value of implementing the alternative.

392