Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TA36

3 - Impact Of Payload Amount On Battery Consumption Rate In a Delivery Application Of Drones Maryam Torabbeigi, University of Houston, UH, Houston, TX, United States, Gino J. Lim, Seon Jin Kim, Navid Ahmadian The drone battery charge limitation is an important factor in drone scheduling in order not to run out of battery during the flight. This study investigates the relationship between battery consumption rate (BCR) and the payload amount, and also the impact of payload amount (customer’s demand) on the drone scheduling. The collected data verifies a linear relationship between BCR and the payload amount. A routing problem is proposed for the drone scheduling. The model determines the number of drones, their path, the assigned customers, and the battery charge at each flight segment. The results show the impact of including BCR in the scheduling. 4 - Multiobjective Uav Route Planning In Continuous Terrain Using a Preference-based Evolutionary Algorithm Murat Mustafa Koksalan, Middle East Technical University, Industrial Engineering Department, Ankara, 06531, Turkey, Erdi Dasdemir, Diclehan Tezcaner Ozturk Multiobjective route planning for unmanned air vehicles (UAV) in continuous terrain involves determining the visiting order of targets and the trajectories used between target pairs under multiple objectives. In this research, a hybrid heuristic approach is developed. Order of targets are first determined with a preference- based evolutionary algorithm converging to the desired regions of the Pareto-optimal frontier using the decision maker’s preferences. Then the trajectories between target pairs are found with a heuristic approach using the results of the evolutionary algorithm. The algorithm is implemented on several problems and results are promising. n TA36 North Bldg 224B Joint Session Drones/Practice Curated: Drones in Logistics Emerging Topic: Robotics, Drones and Autonomous Vehicles in Logistics Emerging Topic Session Chair: James F. Campbell, University of Missouri-St Louis, Saint Louis, MO, 63121-4499, United States 1 - Coordinated Logistics with a Truck and a Drone John Gunnar Carlsson, University of Southern California, 3750 McClintock Avenue, Los Angeles, CA, 90089, United States We determine the efficiency of a delivery system in which an unmanned aerial vehicle (UAV) provides service to customers while making return trips to a truck that is itself moving. In other words, a UAV picks up a package from the truck (which continues on its route), and after delivering the package, the UAV returns to the truck to pick up the next package. By combining a theoretical analysis in the Euclidean plane with real-time numerical simulations on a road network, we demonstrate that the improvement in efficiency is related to the square root of the ratio of the speeds of the truck and the UAV. 2 - Optimization of a Drone-aided Network Bahar Kara, Bilkent University, Department of Industrial Eng, Ankara, 06800, Turkey, Aysu Ozel, Oya Ekin Karasan Integrating drones into delivery networks has advantages such as reduced delivery times, costs and access to hardly reachable points. However, due to the limited abilities of drones, it is not possible to deploy solely drones in delivery networks. Thus, drones should be collaborated with the traditional delivery vehicles and this collaboration requires synchronization between drones and delivery vehicles. In this study we propose a mixed integer mathematical model minimizing the time of the last delivery in a network in which a drone and a truck works in synchronization. 3 - Mothership and Drone Routing Problems Bruce L. Golden, University of Maryland-College Park, 10375 Eclipse Way, Columbia, MD, 21044, United States, Stefan Poikonen The Mothership and Drone Routing Problem considers a tandem between a ship and a drone. The drone is required to visit each of a set of targets. However, the drone has finite battery life and, thus, must coordinate with the ship. The problem combines elements of combinatorial optimization and continuous optimization. Second order cone programming is used in a proposed solution method that is flexible to adapt to alternative objective functions and constraint sets. We then consider several generalizations.

n TA37 North Bldg 225A Data Science and Applied Probability Sponsored: Applied Probability Sponsored Session Chair: Gah-Yi Ban, London Business School, London Business School, London, NW1 4SA, United Kingdom 1 - Statistical Inference for Model Parameters with Stochastic Gradient Descent Xi Chen, New York University, New York, NY, 10012, United States, Jason Lee, Xin Tong, Yichen Zhang In this talk, we investigate the problem of statistical inference of model parameters based on stochastic gradient descent (SGD). To this end, we propose a consistent estimator of the asymptotic covariance of the average iterate from SGD —- batch-means estimator, which only uses the iterates from SGD. As the SGD process forms a time-inhomogeneous Markov chain, our batch-means estimator with carefully chosen increasing batch sizes generalizes the classical batch-means estimator designed for time-homogenous Markov chains. The proposed batch- means estimator allows us to construct asymptotically exact confidence intervals and hypothesis tests. 2 - Assessing the Spillover Effect on Delivery Time: An Empirical Study on a Logistic Platform Yifan Feng, The University of Chicago Booth School of Business, 5807 S. Woodlawn Avenue, Chicago, IL, 60637, United States, Rene A. Caldentey, Linwei Xin Our study is based on a dataset from a major logistics platform for e-commerce business in China. We quantify the spillover effects on delivery time and demonstrate its relationship with the philosophy of complete resource pooling (CRP). 3 - Validating Optimization Under Uncertainty Henry Lam, Columbia University, 500 W. 120th St., New York, NY, 10027, United States, Huajie Qian Optimization formulations to handle decision-making under uncertain constraints, such as (distributionally) robust optimization, often contain parameters that control the level of conservativeness. We investigate strategies to select parameter values by validating their performances in terms of both feasibility and optimality. We demonstrate the effectiveness of these strategies in relation to the optimization class and problem dimension. 4 - Confidence Intervals for Data-driven Inventory Policies with Demand Censoring Gah-Yi Ban, London Business School, Regent’s Park, London, NW1 4SA, United Kingdom We revisit the classical dynamic inventory management problem of Scarf (1959) from the perspective of having n historical selling seasons of data and making ordering decisions for the upcoming season. We develop a nonparametric estimation procedure for the (S; s) policy that is consistent, then characterize the finite-sample properties of the estimated (S; s) levels by deriving their asymptotic confidence intervals. We also consider having at least some of the past selling seasons of data censored from the absence of backlogging. We then show how to correctly use the censored data to obtain consistent decisions and derive asymptotic confidence intervals for this policy using Stein’s method. Joint Session APS/Practice Curated: Bandits, Optimal Stopping, and Control Sponsored: Applied Probability Sponsored Session Chair: David Goldberg, Cornell University, Ithaca, NY, 14850, United States 1 - Recent Progress at the Intersection of Optimal Stopping and Bandits Yilun Chen, Cornell, 292 Rhodes Hall, Ithaca, NY, 14853, United States, David Goldberg We present progress on the problem of optimal stopping, and its application to bandits. We develop a new pure-dual algorithm for optimal stopping, which leads to an expansion (for the optimal value) in which each term has a natural representation in terms of certain (infima of) conditional expectations, and can enable better approximation with fewer nested conditional expectations. For Bayesian bandits, we present several novel insights into the (optimal) Gittins index policy, shedding light on what, qualitatively, makes the Gittins index policy truly optimal, as compared to other policies optimal only w.r.t. (asymptotic) first- order regret. n TA38 North Bldg 225B

256