Skip to Main content Skip to Navigation

Méthodes heuristiques pour le problème de placement sur bande en deux dimensions

Abstract : Packing problems are usually NP-hard, or NP-complete according to the objective. One has to locate a set of objects into one or more “container(s)”, with fix dimensions or of infinite height, while respecting constraints related to some characteristics (weight, quantity, rotation, stability, guillotine cuts...). Themain interest of these problems are the numerous practical applications from various domains. The most effective solution strategies for these problems are usually approximate methods, local search in particular. In this thesis, we are interested in a particular two-dimensional packing problem (without rotation nor guillotine cuts) known as “strip packing” (SPP). The objective of this problem, after locating rectangular objects without overlap, is to minimize the height of the resulting packing. We developed two “meta-heuristic” approaches for the SPP, both including innovative components based on problem knowledge. The first one is a genetic algorithm with a new (highly “visual”) crossover and a hierarchical fitness function. The second one is a tabu search with “direct” representation (i.e. not using the classical permutations) whose main characteristics are a consistent neighborhood, a “well-informed” diversification (based on the search history), and a fitness function able to evaluate possibly partial solutions. The two proposed approaches, assessed on a well-known and very difficult benchmark, show good performances compared with other strategies.
Document type :
Complete list of metadatas

Cited literature [79 references]  Display  Hide  Download
Contributor : Anne-Marie Plé <>
Submitted on : Friday, March 11, 2011 - 11:43:41 AM
Last modification on : Monday, June 1, 2020 - 4:54:02 PM
Long-term archiving on: : Sunday, June 12, 2011 - 2:30:10 AM


  • HAL Id : tel-00575859, version 1


Giglia Gomez-Villouta. Méthodes heuristiques pour le problème de placement sur bande en deux dimensions. Informatique [cs]. Université d'Angers, 2010. Français. ⟨tel-00575859⟩



Record views


Files downloads