Journées de l'optimisation 2022

HEC Montréal, Québec, Canada, 16 — 18 mai 2022

Horaire Auteurs Mon horaire

WC3 - Industrial session: IVADO Labs and Artelys

18 mai 2022 15h30 – 17h10

Salle: Procter & Gamble (vert)

Présidée par Marie-Claude Côté

5 présentations

  • 15h30 - 15h55

    Store specific planogram customization at a large national Canadian retailer

    • Adrien Rimélé, prés., Ivado Labs
    • Louis-Philippe Bigras, Ivado Labs
    • Mouad Faik, Ivado Labs
    • Philippe Grangier, Ivado Labs
    • Nabila Remli, Ivado Labs
    • Yossiri Adulyasak, HEC Montréal
    • Andrea Lodi, Polytechnique Montreal
    • Maxime Cohen, McGill university
    • Louis-Martin Rousseau, Polytechnique Montréal

    In retail, a planogram is a plan indicating which products and which fixtures should be placed where, within an aisle. These are typically created by corporate experts at the national level, and then require some adaptation by employees in each store to account for local variations (assortment, aisle geometry, forecast…). This adjustment process is time consuming and can lead to suboptimal store decisions. We present a tool developed in partnership between IVADO Labs and a large national Canadian retailer to automatically generate store specific planograms.

  • 15h55 - 16h20

    Solving a real-life pickup and delivery problem with supply-demand balancing

    • Camilo Ortiz-Astorquiza, prés., Ivado Labs
    • Caroline Rocha, Ivado Labs
    • Jean-François Cordeau, HEC Montréal, GERAD, CIRRELT
    • Guy Desaulniers, GERAD - Polytechnique Montréal
    • Jorge Mendoza, HEC Montréal

    In this talk, we present a real-life application of a vehicle routing problem with various side constraints in the context of waste and supply management for water treatment facilities in North America, pointing out the main features and challenges encountered in practice. In general terms, for a rolling planning horizon of five days, given a set of steel mills which produce a waste material, and a set of water treatment facilities that consume such product, the problem consists of finding the set of routes with their assignment of resources (trailers, trucks and drivers) so that the waste product of the steel mills is delivered to the customers in a timely manner while minimizing transportation cost. We also present the solution approach that has been implemented and discuss the preliminary computational results obtained with the current instances.

  • 16h20 - 16h45

    Stochastic Storage Allocation Problem at a Container Transloading Terminal

    • Nicolas Boez, prés., Ivado Labs
    • Simon Boivin, Ray-Mont Logistics
    • Louis-Philippe Bigras, Ivado Labs
    • Jean-François Cordeau, HEC Montréal, GERAD, CIRRELT
    • Andrea Lodi, Polytechnique Montreal

    Ray-Mont Logistics, a company specialized in international freight transportation, worked with Ivado Labs to improve the management of containers in their transloading terminals. The aim was to determine the optimal location of new loaded containers to reduce retrieval efforts when shipping out containers. The departure date of a container may change after it is stacked, therefore the uncertainty of this information had to be considered to provide robust decisions. The solution, involving a mixed-integer program associated with a Monte Carlo simulation, allowed to significantly reduce operational efforts by almost eliminating “buried” containers in most situations, especially when departure dates shift.

  • 16h45 - 17h10

    Power flow optimization using Artelys Knitro

    • Carlyle Deligny, prés.,

    Introduction of Artelys Knitro
    At Artelys, we specialize in optimization, decision-support, modeling, and delivering efficient solutions to complex business problems. Artelys Knitro is the industry-leading solver for very large nonlinear optimization problems (NLP, MINLP) used by more than 450 institutions in over 40 countries. The first part of this session will focus on introducing Artelys Knitro and its features.

    Artelys Knitro 13
    Mixed-Integer Nonlinear Problems (MINLP) arise in a wide range of fields (e.g engineering, finance, power networks, etc.) Released beginning 2022, Artelys Knitro 13 introduces a new parallel branch-and-bound algorithm to solve complex MINLP problems and is also relying on 4 threads on hard MINLP instances to reduce its computation time. The upgrade also comes with performance improvements on hard continuous nonlinear problems and ARM architectures.

    Use case
    The GO competition is a challenge focusing on security-constrained optimal power flow (SCOPF) and involves developing power system optimization and control algorithms on synthetic and real network models. In 2021, Challenge 2 concentrated on the topology optimization problem and expanded upon the SCOPF problem. Artelys team developed a transformational way to solve power flow optimization problems using Artelys Knitro.

  • 17h10 - 17h35

    Introduction of Artelys solvers – market clearing use case

    • Julien Côté-Massicotte, prés., Artelys
    • Carlyle Deligny, Artelys

    1. Artelys introduction and its solvers
    Artelys is specialized in optimization, decision-support and modeling.
    - Xpress: solver specialized in linear and mixed-integer problems
    - Knitro: Nonlinear programming
    - AMPL: a modeling language for linear and nonlinear optimization problems
    - Artelys Kalis: Constraints programming component able to solve combinatorial problems

    2. Use case : The European Market clearing platform TERRE

    Artelys supports the design, development and implementation of energy market software by leveraging its skills in mathematical optimization and its solvers. TERRE (Trans European Replacement Reserve Exchange) project delivers a European platform for the exchange of balancing energy from replacement reserves based on LIBRA solution, a common IT system which supports the exchange of the balancing energy.
    The challenge was to develop a market clearing algorithm to select the optimal set of bids to cover the replacement reserve needs and ensure that the solutions respect complex business rules. The problem has been modelled and solved with FICO® Xpress Optimization Suite. Artelys and FICO have been working together for over 20 years, providing optimization services and solutions to a diverse range of industries.