Showing papers in "ERIM report series research in management Erasmus Research Institute of Management in 2016"

Crowdsourced Delivery: A Dynamic Pickup and Delivery Problem with Ad-hoc Drivers

Abstract: The trend towards shorter delivery lead-times reduces operational efficiency and increases transportation costs for internet retailers. Mobile technology, however, creates new opportunities to organize the last-mile. In this paper, we study the concept of crowdsourced delivery that aims to use excess capacity on journeys that already take place to make deliveries. We consider a peer-to-peer platform that automatically creates matches between parcel delivery tasks and ad-hoc drivers. The platform also operates a fleet of backup vehicles to serve the tasks that cannot be served by the ad-hoc drivers. The matching of tasks, drivers and backup vehicles gives rise to a new variant of the dynamic pick-up and delivery problem. We propose a rolling horizon framework and develop an exact solution approach to solve the various subproblems. In order to investigate the potential benefit of crowdsourced delivery, we conduct a wide range of computational experiments. The experiments provide insights into the viability of crowdsourced delivery under various assumptions about the environment and the behavior of the ad-hoc drivers. The results suggest that the use of ad-hoc drivers has the potential to make the last-mile more cost-efficient and can reduce the system-wide vehicle-miles.

Model-based Purchase Predictions for Large Assortments

Abstract: Being able to accurately predict what a customer will purchase next is of paramount importance to successful online retailing. In practice, customer purchase history data is readily available to make such predictions, sometimes complemented with customer characteristics. Given the large assortments maintained by online retail- ers, scalability of the prediction method is just as important as its accuracy. We study two classes of models that use such data to predict what a customer will buy next: A novel approach that uses latent Dirichlet allocation (LDA), and mixtures of Dirichlet-Multinomials (MDM). A key benefit of a model-based approach is the potential to accommodate observed customer heterogeneity through the inclusion of predictor variables. We show that LDA can be extended in this direction while retaining its scalability. We apply the models to purchase data from an online re- tailer and contrast their predictive performance with that of a collaborative filter and a discrete choice model. Both LDA and MDM outperform the other meth- ods. Moreover, LDA attains performance similar to that of MDM while being far more scalable, rendering it a promising approach to purchase prediction in large assortments.

Optimization Approaches for the Traveling Salesman Problem with Drone

Abstract: textThe fast and cost-efficient home delivery of goods ordered online is logistically challenging. Many companies are looking for new ways to cross the last-mile to their customers. One technology-enabled opportunity that recently has received much at- tention is the use of a drone to support deliveries. An innovative last-mile delivery concept in which a truck collaborates with a drone to make deliveries gives rise to a new variant of the traveling salesman problem (TSP) that we call the TSP with drone. In this paper, we model this problem as an IP and develop several fast route first-cluster second heuristics based on local search and dynamic programming. We prove worst-case approximation ratios for the heuristics and test their performance by comparing the solutions to the optimal solutions for small instances. In addition, we apply our heuristics to several artificial instances with different characteristics and sizes. Our experiments show that substantial savings are possible with this concept in comparison to truck-only delivery.

Evaluating dedicated and shared storage policies in robot-based compact storage and retrieval systems

Abstract: textRobot-based compact storage and retrieval systems (RCSRS) have seen many implementations over the last few years. In such a system, the inventory items are stored in bins, organized in a grid. In each cell of the grid, a certain number of bins are stored on top of each other. Robots with transport and lifting capabilities move on the grid roof to transport bins between manual workstations and storage stacks. We estimate performance and evaluate storage policies of RCSRS, considering both dedicated and shared storage policies coupled with random and zoned storage stacks. Semi-open queueing networks (SOQNs) are built to estimate the system performance, which can handle both immediate and delayed reshuffling processes. We approximate the models by reduced SOQNs with two load-dependent service nodes and use the Matrix-Geometric Method (MGM) to solve them. Both simulations and a real case are used to validate the analytical models. Assuming a given number of stored products, our models can be used to optimize not only the length to width ratio of the system, but also the stack height, depending on the storage strategy used. For a given inventory and optimal system configuration, we demonstrate that the dedicated storage policy outperforms the shared storage policy in terms of dual command throughput time. However, from a cost perspective, with a maximum dual command throughput time as a constraint, we show that shared storage substantially outperforms dedicated storage. The annualized costs of dedicated storage are up to twice as large as those of shared storage, due to the larger number of storage positions required by dedicated storage and the relatively lower filling degree of storage stacks.

The 2016 Power Trading Agent Competition

Abstract: markdownThis is the specification for the Power Trading Agent Competition for 2016 (Power TAC 2016). Power TAC is a competitive simulation that models a “liberalized” retail electrical energy market, where competing business entities or “brokers” offer energy services to customers through tariff contracts, and must then serve those customers by trading in a wholesale market. Brokers are challenged to maximize their profits by buying and selling energy in the wholesale and retail markets, subject to fixed costs and constraints; the winner of an individual “game” is the broker with the highest bank balance at the end of a simulation run. Costs include fees for publication and withdrawal of tariffs, and distribution fees for transporting energy to their contracted customers. Costs are also incurred whenever there is an imbalance between a broker’s total contracted energy supply and demand within a given time slot. The simulation environment models a wholesale market, a regulated distribution utility, and a population of energy customers, situated in a real location on Earth during a specific period for which weather data is available. The wholesale market is a relatively simple call market, similar to many existing wholesale electric power markets, such as Nord Pool in Scandinavia or FERC markets in North America, but unlike the FERC markets we are modeling a single region, and therefore we approximate locational-marginal pricing through a simple manipulation of the wholesale supply curve. Customer models include households, electric vehicles, and a variety of commercial and industrial entities, many of which have production capacity such as solar panels or wind turbines. All have “real-time” metering to support allocation of their hourly supply and demand to their subscribed brokers, and all are approximate utility maximizers with respect to tariff selection, although the factors making up their utility functions may include aversion to change and complexity that can retard uptake of marginally better tariff offers. The distribution utility models the regulated natural monopoly that owns the regional distribution network, and is responsible for maintenance of its infrastructure. Real-time balancing of supply and demand is managed by a market-based mechanism that uses economic incentives to encourage brokers to achieve balance within their portfolios of tariff subscribers and wholesale market posi- tions, in the face of stochastic customer behaviors and weather-dependent renewable energy sources. Changes for 2016 are focused on a more realistic cost model for brokers, and are highlighted by change bars in the margins. See Section 7 for details.

Maintenance Appointments in Railway Rolling Stock Rescheduling

Abstract: This paper addresses the Rolling Stock Rescheduling Problem (RSRP), while taking maintenance appointments into account. After a disruption, the rolling stock of the disrupted passenger trains has to be rescheduled in order to restore a feasible rolling stock circulation. Usually, a number of train units have a scheduled maintenance appointment during the day: these appointments must be taken into account while rescheduling the rolling stock. In this paper we propose three Mixed Integer Programming (MIP) models for this purpose. All models are extensions of the Composition model from literature, which does not distinguish individual train units. The Extra Unit Type model adds an additional rolling stock type for each train unit that requires maintenance. The Shadow- Account model keeps track of a shadow account for each train unit that requires maintenance. The Job-Composition model creates a path for each train unit such that the train units that require maintenance are on time for their maintenance appointments. All models are tested on instances of Netherlands Railways (NS). The results show that especially the Shadow-Account model and the Job-Composition model are effectively able to take maintenance appointments into account during real-time rescheduling. It depends on the characteristics of an instance whether the Shadow-Account model or the Job-Composition model performs best.

Vertical or Horizontal Transport? - Comparison of robotic storage and retrieval systems

Abstract: textAutonomous vehicle-based storage and retrieval systems are commonly used in e-commerce fulfillment as they allow a high and flexible throughput capacity. In these systems, roaming robots transport loads between a storage location and a workstation. Two main variants exist: Horizontal, where the robots only move horizontally and use lifts for vertical transport and a new variant Vertical, where the robots can also travel vertically in the rack. This paper builds a framework to analyze the performance of the vertical system and to compare its throughput capacity with the horizontal system. We build closed-queueing network models for this that in turn are used to optimize the design. The results show that the optimal height-to-width ratio of a vertical system is around 1. As a large number of system robots may lead to blocking and delays, we compare the effect of two different robot blocking protocols on the system throughput: robot Recirculation and Wait-On-Spot. The Wait-On-Spot policy produces a higher system throughput when the number of robots in the system is small. However, for a large number of robots in the system, the Recirculation policy dominates the Wait-On-Spot policy. Finally, we compare the operational costs of the vertical and the horizontal transport system. For systems with one load/unload (L/U) point, the vertical system always produces a similar or higher system throughput, with a lower operating cost compared to the horizontal system with a discrete lift. It also outperforms the horizontal system with a continuous lift in systems with two L/U points.

Competitive Capacity Investment under Uncertainty

Abstract: We consider a long-term capacity investment problem in a competitive market under demand uncertainty. Two firms move sequentially in the competition and a firm’s capacity decision interacts with the other firm’s current and future capacity. Throughout the investment race, a firm can either choose to plan its investments proactively, taking into account possible responses from the other firm, or decide to respond reactively to the competition. In both cases, the optimal decision at each period is determined according to an ISD (Invest, Stayput, Disinvest) policy. We develop two algorithms to efficiently derive proactive ISD policies for the leader and follower firms. Using data from the container shipping market (2000-2015), we show that the optimal capacity determined by our competitive strategy is consistent with the realized investments in practice. By revealing strategical flexibility of proactive strategies, our results demonstrate that firms in the competition can gain more capacity and profit through such a strategy. Using Monte Carlo simulations, we explore the impact of different market conditions and investment irreversibility levels on capacity strategies. In particular, by comparing the results of competitive strategies and strategies that separate firms into different markets, we show that both firms can benefit from the competition and that market downturns likely lead to investment cascades.

Demand Heterogeneity and the Adoption of Platform Complements

Abstract: textThis paper offers a demand-based theory of how platform maturity affects the adoption of platform complements. We argue that differences between early and late adopters of the platform include willingness to pay for the platform-and-complement bundle, risk preferences, preference for novelty, and search behavior. These differences create heterogeneous demand conditions for complements that affect both average complement performance and variance in the types of complements that are more or less successful. Using a novel dataset of 2,921 sixth-generation console video games, we find that platform maturity has a negative relationship on video games’ unit sales. Furthermore, as the platform matures, we find that the sales disparity between new intellectual property (IP) games and games based on existing video game properties or media tie-ins grows to the detriment of new IP games. We find that the sales disparity between superstar games and flops also widens as the platform matures. These effects are accentuated by the introduction of a next generation platform, which further skews the complement’s customer pool as early adopters migrate away from the current generation platform. Robustness tests that control for unobserved heterogeneity help rule out alternative explanations and support our argument that these performance implications are truly driven by heterogeneity in demand.