4 Presentations

• 
  03:30 PM - 03:55 PM

  Improved branch-and-cut algorithm for the inventory routing problem

  • Jørgen Skålnes, presenter, Norwegian University of Science and Technology
  • Magnus Stålhane, Norwegian University of Science and Technology
  • Henrik Andersson, Norwegian University of Science and Technology
  • Guy Desaulniers, GERAD - Polytechnique Montréal

  Combining the valid inequalities from the current state-of-the-art branch-and-cut algorithm and branch-and-price algorithm together with a new concept, delivery patterns, we are able to obtain stronger dual bounds in the root node and obtain better best known solutions for the benchmark instances on the Inventory Routing Problem.

  Keywords: Branch-and-Cut, Valid Inequalities, IRP
• 
  03:55 PM - 04:20 PM

  Distribution network design and inventory planning under uncertainty

  • Amira Dems, presenter, Polytechnique Montréal
  • Jean-François Cordeau, HEC Montréal, GERAD, CIRRELT
  • Yossiri Adulyasak, HEC Montréal

  This paper deals with an integrated planning problem where the distribution network design and inventory planning decisions are optimized simultaneously under demand uncertainty. Since the level of safety stocks to guarantee required service levels depends essentially on the supply lead time and the variability of the pooled demand, which are the result of the supply network decisions, the integrated tool allows the planner to take into account this complex relationship and determine an optimal plan. To tackle this problem, first, we present an exact solution approach based on a branch-and-check framework and propose a new set of valid cuts to solve the general case when the demand can take any form. This approach, albeit general, is not highly scalable, thus we present a computational enhancement using inequalities derived from piecewise linear functions. We consider the case where the demand is represented by a normal distribution as well as a more general setting where the demand is irregular and represented by a negative binomial distribution. We present extensive computational results on datasets adapted from the literature to validate the proposed modeling and solution approaches. Then, we conduct several tests to study the sensitivity of the network design decisions to different cost and capacity scenarios. Finally, we evaluate the integrated solution against the sequential one, where the distribution
  network design and the inventory decisions are made separately.

  Keywords: Inventory-location, branch-and-check, stochastic demand, distribution network design, inventory planning, logic-based Benders decomposition.
• 
  04:20 PM - 04:45 PM

  Benders decomposition for a three-level lot sizing and replenishment problem.

  • Matthieu Gruson, presenter, HEC Montréal/CIRRELT
  • Jean-François Cordeau, HEC Montréal, GERAD, CIRRELT
  • Raf Jans, HEC Montréal

  We address a three-level lot sizing and replenishment problem with a distribution structure (3LSPD). We use a Benders decomposition approach and develop a branch-and-cut algorithm to efficiently solve both a deterministic and a stochastic variant of the uncapacitated case of the problem. The algorithm we propose outperforms CPLEX.
• 
  04:45 PM - 05:10 PM

  Inventory management using an (s,S)-policy at the Bank of Canada

  • Antoine Legrain, presenter, Polytechnique Montréal
  • Jonathan Patrick, Telfer School of Management, University of Ottawa

  The Bank of Canada (BoC) has 43 Regional Distribution Centers (RDC) to ensure that there is sufficient inventory of each denomination of currency all across the country. Local banks can withdraw from the RDCs or deposit money into the RDCs. Thus the BoC is faced with a two-way inventory management problem where demand can be both positive (withdrawals) or negative (deposits) and where the Bank must avoid shortages and overcaps. We developed an adapted (s,S)-policy that has been implemented by the BoC through 2018 and has demonstrated a 20% drop in transportation costs compared to the previous years.