Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA33

Due to expansion of online shopping market, online store is increasing. Many companies are focusing on improving delivery as it leads to greater customer satisfaction and reduced delivery costs. In many companies, delivery of online stores is delivered with shifted time window by several hours from the time window, in addition to the conventional time window, in order to improve the usage rate of delivery docks. This is called an overlapped time window. In this paper, we aim to contribute to the delivery planning of online store by proposing a model of delivery planning considering overlapped time window and approximate solution using simulated annealing. 2 - Solving Real World Vehicle Routing Problems at Scale Bhanu Krishna Potluri, Llamasoft Inc., Ann Arbor, MI, 48104, United States Vehicle routing problems are known to be notoriously difficult to solve. Modeling real-world constraints such as pickup-delivery time windows, driver hours of service, fleet sizing, etc. further increases complexity. With expanding logistics networks and ever increasing global freight volumes, businesses want to solve problems of realistic scale with real world constraints. This session explores meta- heuristics and cloud computing to achieve scalability of vehicle routing algorithms. 3 - Allowing for Re-optimisation in the Vehicle Routing Problem with Time Windows, Stochastic Customers and Stochastic Demand: Model and Solution Methods Vincentius Cornelis Gerardus Karels, PhD, Technical University Eindhoven, Duikerstraat 12 Bis, Utrecht, 3582 TB, Netherlands, Lucas Petrus Veelenturf, Tom Van Woensel We solve a vehicle routing problem with both deterministic and stochastic customers and stochastic demands. For the deterministic customers furthermore self-imposed time-windows are determined. The problem is modeled as a two- stage stochastic programming model. Whereas most similar problems allow for simple recourse in the second stage, we allow for re-optimisation. This introduces additional complexity in the model, for which we will introduce novel solution methods. 4 - Multi-depot Electric Vehicle Routing Problem with Time Windows Considering Non-linear Charging and Discharging Surendra Reddy Kancharla, Indian Institute of Technology Madras, #238, Building Sciences Block, Alumni Ave, Chennai, 600036, India, Gitakrishnan Ramadurai A new variant of Electric Vehicle Routing Problem with Time Windows (EVRPTW) to find the routes and charging schedules of vehicles operating from multiple depots with the objective of minimizing energy consumption is proposed. The present variant does not limit the number of visits to a charging station and considers non-linear charging and discharging. A mixed-integer program is formulated and an effective heuristic solution algorithm is presented. 5 - A Vehicle Routing Problem with Drones Amro El-Adle, University of Massachusetts Amherst, 121 Presidents Drive, Amherst, MA, 01003, United States, Mohammad Reihaneh, Ahmed Ghoniem We investigate a vehicle routing problem with drones (VRPD) in which customers may be served either by delivery vehicles or by unmanned aerial drones launched from the vehicles. Building on the success of branch-and-price (BP) algorithms for vehicle routing problems, we use a labeling algorithm that generates synchronized drone and vehicle routes. We demonstrate the usefulness of the proposed methodology in our computational study. n WA33 North Bldg 222C Inventory Management I Contributed Session Chair: Shunichi Ohmori, Waseda University, Room 0903A, Okubo 3-4-1, Shinjuku, Tokyo, Japan 1 - Determination of Order Amounts for Multiple Spare Parts of Products in the Post-production Phase Nicholas William Leifker, Associate Professor, St. John Fisher College, 3690 East Ave., Rochester, NY, 14618, United States, Timothy Joe Lowe, Philip C. Jones Manufacturers often encounter difficulties in supplying an adequate number of spare parts for a product that is in its post-production phase. As a result, manufacturers will sometimes make one final order of spare parts, which is used to satisfy any demand for spare parts going forward. This order may be complicated if the failure rate for one type of spare part is partially dependent on the demand for other types of spare parts, if the failure rate of the product is partially dependent of the failure rates of its parts, or if it becomes advantageous for a spare part to be salvaged. We compare two solution methods to find the order amount: a dynamic programming method, and an iterative optimization method.

n WA31 North Bldg 222A Transportation-Planning I Contributed Session Chair: Albert Schrotenboer, University of Groningen, Vlasstraat 19, Groningen, 7912KS, Netherlands 1 - A Simulation Approach to the Stochastic Vehicle Routing Without Fitting Demands We propose a method to solve the Stochastic Vehicle Routing Problem (SVRP) with stochastic demands. We realize that the distributional information of customer demands is, at best, empirical data only and the real-life SVRPs would be in a very difficult form. In this paper, we discuss the similarity of different demand scenarios and remove many scenarios, which are quite similar to each other. By greatly reducing the number of scenarios, we build a tractable SVRP model with a large number of demand scenarios considered. We then compared our model with two other models: the metaheuristic of vehicle routing, and the vehicle routing model with two-stage recourse. 2 - Freight Mode Choice: A Machine Learning Approach Rodrigo Mesa Arango, Florida Institute of Technology, Melbourne, FL, 32901, United States Presents a machine learning application for mode choice prediction in freight transportation planning. Discusses model training through algorithm selection, model parameter optimization, and over-fitting avoidance. Contrasts challenges and opportunities related to classic econometric approaches. Employs 4.5 million shipments available in the 2012 Commodity Flow Survey Public Use Microdata for model estimation/validation. 3 - Proactive Medical Transportation Planning Mohammed Skiredj, PhD Student, École Nationale Supérieure des Mines de Saint-Étienne, 158 cours Fauriel, Saint-Étienne, 42023, France, Xiaolan Xie, Thierry Garaix This work treats a real-life health-care transportation problem for Non- Emergency ambulance service. The problem can be seen as a multi-trip pickup and delivery problem with time windows and Trip-Crew-Vehicle assignment. the objective is to serves as many patients as possible, minimizes the total distance traveled by vehicles, and finally preserves as far as possible the forecaster starting and ending time of each driver. To solve the problem we proposed an adaptive large neighborhood search metaheuristic that uses a constructive greedy heuristic to build the initial solution. 4 - Data-driven Methodology for Bicycle use Propensity Estimation Pablo Andr s Uriza-Antorveza, Universidad de los Andes, Edificio Mario Laserna Cra 1 Este No 19A - 40, Bogotá, 10101, Colombia, Sergio Cabrales, Andres L. Medaglia, Olga L. Sarmiento As part of adopting greener and healthier lifestyles, local governments are promoting bicycle use by implementing policies linked to the built environment. As part of a larger analytics framework designed to support transit planners, we developed a bicycle-use propensity model that lies at the core of the decision support system and feeds from geospatial data from different sources. We illustrate how the proposed method is able to support strategic decisions by quantifying the impact of a policy with data from Bogotß (Colombia). 5 - Robust Fleet Size and Mix at Offshore Wind Farms Albert Schrotenboer, PhD Candidate, University of Groningen, Groningen, Netherlands, Evrim Ursavas, Iris F. Vis We introduce the Robust Fleet Size and Mix Problem (RFSMP) inspired by vehicle chartering decisions at offshore wind farms. A novel multi-commodity flow formulation is proposed and reformulated by means of a Dantzig-Wolfe decomposition into a set partitioning-like formulation. This leads to a master problem of which both the number of rows and the number of columns is extremely large. In addition, the number of subproblems depends on the number of included scenarios. In this talk, we discuss an exact approach based on column generation that is capable of solving the RFSMP. n WA32 North Bldg 222B Practice- Vehicle Routing I Contributed Session Chair: Amro El-Adle, University of Massachusetts, 121 Presidents Drive, Amherst, MA, 01003, United States 1 - Vehicle Routing Problem with Overlapped Time Windows on Shipping from Online Store Takashi Irohara, Professor, Sophia University, 7-1 Kioi-cho, Chiyoda-ku, Tokyo, 102-8554, Japan Robert A. Russell, University of Tulsa, Finance & Operations Mgmt Dept, 800 South Tucker Drive, Tulsa, OK, 74104, United States, Wen-Chyuan Chiang, Lijian Chen

421