Time filter

Source Type

Montréal, Canada

Contardo C.,ESG UQAM | Desaulniers G.,Ecole Polytechnique de Montreal | Lessard F.,Groupe dEtudes et de Recherche en Analyse des Decisions GERAD
Networks | Year: 2015

In this article, we present a comparative study of several strategies that can be applied to achieve the so-called elementary lower bound in vehicle routing problems, that is, the bound obtained when all positive-valued variables in an optimal solution of the linear relaxation of the set-partitioning formulation correspond to vehicle routes without cycles. This bound can be achieved by solving the resource-constrained elementary shortest path problem - an NP-hard problem - as the pricing problem in a column generation algorithm, but several other strategies can be used to ultimately produce the same lower bound in less computational effort. State-of-theart algorithms for vehicle routing problems rely on the quality of this lower bound to either bound the size of the search tree in a branch-and-price algorithm or the complexity of an enumeration procedure used to limit the number of variables in the set-partitioning model. We consider several strategies for imposing elementarity that involve ng -paths, strong degree constraints, and decremental state-space relaxation. We compare the performance of these strategies on some selected instances of the vehicle routing problem with time windows. © 2015 Wiley Periodicals, Inc. Source

Contardo C.,ESG UQAM | Cordeau J.-F.,HEC Montreal | Gendron B.,University of Montreal
Discrete Optimization | Year: 2013

In this paper we present a computational comparison of four different flow formulations for the capacitated location-routing problem. We introduce three new flow formulations for the problem, namely a two-index two-commodity flow formulation, a three-index vehicle-flow formulation and a three-index two-commodity flow formulation. We also consider an existing two-index vehicle-flow formulation and extend it by considering new families of valid inequalities and separation algorithms. We introduce new branch-and-cut algorithms for each of the formulations and compare them on a wide number of instances. Our results show that compact formulations can produce tight gaps and solve many instances quickly, whereas three-index formulations scale better in terms of computing time. © 2013 Elsevier B.V. All rights reserved. Source

Purpose: Knowledge management (KM) has become a key concern for companies which nowadays are constantly looking for better ways to assure knowledge sharing between their employees. However, companies encounter several challenges arising from the fact that several generations share the same workplace and a big portion of today's employees are close to retirement. This article aims to focus on knowledge sharing between generations. Design/methodology/approach: The article reviews the "generation" concept and its limitations, and introduces a new view on generations as "communities of knowledge". Findings: Companies have to find ways not only to assure knowledge transfer between generations, but also knowledge retention of the workers that are retiring. This requires a deeper understanding of the generations and their differentiated knowledge. Yet, today's dominant descriptions of generations ("baby-boomers", "generation X", "generation Y"), do not appear to adequately take into account cultural, socio-professional and individual factors. Originality/value: The proposed change of paradigm allows a deeper comprehension of nuances that may exist within the same age group. In doing so, the article makes a contribution to the understanding of knowledge sharing in organizations. © Emerald Group Publishing Limited. Source

Plouffe G.,Ecole Polytechnique de Montreal | Bulle C.,ESG UQAM | Deschenes L.,Ecole Polytechnique de Montreal
Journal of Cleaner Production | Year: 2015

Metals are a major concern in life cycle assessment (LCA): they dominate the ecotoxicological impacts due to the use of models that do not consider metal speciation. The objective of this study was to measure the influence of zinc (Zn) speciation on LCA results and evaluate the importance of including it in LCA. A case study in which ecosystem quality impacts are dominated by zinc (Zn) when using the current characterization factor (CF) for Zn was performed on the market for electricity, low voltage CA-QC (Canada-Quebec) ecoinvent data using the IMPACT 2002 + methodology. Different CFs for Zn were tested: IMPACT 2002 + CF and USEtox-derived terrestrial CF without speciation, generic default CF values for the world, and minimum and maximum regionalized CF values including speciation based on either soluble or true solution Zn. For each scenario, the following outcomes were determined and compared with those obtained using the other CFs for Zn: the contribution of Zn emitted to soil to the terrestrial ecotoxicity impact category and ecosystem quality damage category scores, the rank of Zn emitted to soil as a contributor to the terrestrial ecotoxicity impact category, and the total terrestrial ecotoxicity impact score. To our knowledge, this is the first study that includes zinc speciation in soil in a real LCA case study. In this case study, when including speciation, the contribution of Zn emitted to soil to the terrestrial ecotoxicity impact score falls from 26% to a maximum of 1.27%; the impact scores of Zn emitted to soil drop by 1.44-14.37 orders of magnitude; the total terrestrial ecotoxicity and ecosystem quality impact scores decrease by approximately 25% and 21%, respectively; and among the major contributors to terrestrial ecotoxicity, Zn falls from the 2nd to between the 9th and 255th position. Considering that the contribution of all metals emitted to soil represents between 56.2% and 75.7% of the total terrestrial ecotoxicity impact score for this case study, the results highlight the need to include metal speciation in LCA and the considerable influence it could have on ecotoxicological impacts. © 2015 Elsevier Ltd. Source

Contardo C.,ESG UQAM | Cordeau J.-F.,HEC Montreal | Gendron B.,University of Montreal
INFORMS Journal on Computing | Year: 2014

In this paper we present an exact algorithm for the capacitated location-routing problem (CLRP) based on cut-and-column generation. The CLRP is formulated as a set-partitioning problem that also inherits all of the known valid inequalities for the flow formulations of the CLRP. We introduce five new families of inequalities that are shown to dominate some of the cuts from the two-index formulation. The problem is solved by column generation, where the subproblem consists in finding a shortest path of minimum reduced cost under capacity constraints. We first use the two-index formulation for enumerating all of the possible subsets of depot locations that could lead to an optimal solution of cost less than or equal to a given upper bound. For each of these subsets, the corresponding multiple depot vehicle routing problem is then solved by means of column generation. The results show that we can improve the bounds found in the literature, solve to optimality some previously open instances, and improve the upper bounds on some other instances. © 2014 INFORMS. Source

Discover hidden collaborations