4 présentations

• 
  10h30 - 10h55

  Reliable Adjustable Planning in the Case of Dynamically Arriving Information: The example of E-Commerce Warehouses

  • Michel Gendreau, prés., Polytechnique Montréal
  • Catherine Lorenz, Chair of Management Science/Operations and Supply Chain Management, University of Passau
  • Alena Otto, Chair of Management Science/Operations and Supply Chain Management, University of Passau

  In the ongoing transition towards Industry 4.0, planners seek to leverage optimization algorithms for well-informed decision-making in dynamic contexts, to solve complex planning problems in real-time. This necessitates the development of intelligent adjustable algorithms, whose behavior is well understood for reliable practical deployment.
  This talk focuses on the application of adjustable planning for picking operations in E-commerce warehouses. Since order picking accounts for a substantial portion (60-70%) of total operating costs in warehouses, it holds a pivotal role. In E-commerce settings, picking operations are essentially online, with customer orders arriving dynamically in real-time, highlighting the need for adjustable algorithms, or online policies.
  Although several online policies for order picking have been suggested in the literature, there is a notable lack of understanding of their performance with respect to optimality. We conduct a comprehensive examination of a prominent online warehousing policy – myopic, immediate reoptimization (Reopt) – and establish analytical performance bounds using probabilistic and worst-case analysis. We demonstrate that, under broad stochastic assumptions, Reopt is almost surely asymptotically optimal and we validate its near-optimal performance empirically.
  Our analysis offers significant insights into long-standing discussions in the warehousing literature about the merits of forward-looking policies that aim to anticipate future orders.

• 
  10h55 - 11h20

  Duality in Optimal Stopping

  • Nicolas Gagné, prés., Université de Montréal
  • Fabian Bastin, Université de Montréal

  Optimal stopping problems, commonly encountered in financial mathematics, involve determining the most opportune time to execute an action (e.g., exercising an option) to maximize expected returns. We introduce the dual problem associated to an optimal stopping problem by way of a perturbation function. We give an intuitive interpretation of the dual problem as well as the role it plays in practice.

• 
  11h20 - 11h45

  A stochastic programming approach for production planning in smart factories

  • Benedikt Kern, prés., Concordia University
  • Masoumeh Kazemi Zanjani, Concordia University

  In this study, we propose a two-stage stochastic programming model to address operational-level production planning challenges within a smart factory. Specifically, we focus on a modular production setup within a convertible final production facility, aiming to produce highly customizable modular-structured products. Our model incorporates two sources of uncertainty: random fluctuations in product demand and a stochastic machine degradation affecting production capacity. The proposed stochastic mixed-inter programming (SMIP) optimizes machine activations and production volumes with the objective of minimizing total costs. To manage uncertainty, we introduce recourse actions that activate additional machines and schedule maintenance to maintain adequate capacity. To overcome the computational challenge of the SMIP model, we apply a scenario decomposition approach coupled with benders decomposition. To validate our approach, we conduct extensive computational experiments and compare its performance to that of a commercial solver.
  Keywords – Operational production planning, smart factory, stochastic mixed-integer programming, scenario decomposition, benders decomposition

• 
  11h45 - 12h10

  Procurement and Production Planning with Stochastic Demands

  • Paziani Tomazella Caio, HEC Montreal
  • Oliveira Santos Maristela, University of São Paulo
  • Alem Douglas, University of Edinburgh
  • Raf Jans, prés., HEC Montréal

  We consider a two-level production planning problem. The problem is modeled as a two-level lot sizing problem over a discrete time horizon. One level considers the production decision of final items, taking into account limited production capacity and the other level considers the procurement decision, including which supplier to buy from in each period. Since the demands are uncertain, the problem is modeled as a two-stage stochastic programming model. We analyse different levels of production flexibility, and propose an adaptive Sample Average Approximation algorithm to solve the problem.