Skip to Main content Skip to Navigation
Theses

Outils d'aide à la décision pour des problèmes d'ordonnancement dynamiques

Abstract : Scheduling problems represent an important class of combinatorial optimization problems. Most work in this area focus on static problems in which all data (activities, resources, constraints) are known in advance. In practice, however, schedules are subject to unexpected events (late arrival of raw materials, new orders, machine breakdowns, etc.). Therefore, a computed schedule rarely follows the original plan. It is important to deal with dynamic scheduling problems. In this thesis, we consider a general scheduling problem: the Resource Constrained Project Scheduling Problem (RCPSP for short), and we propose a system that is able to solve dynamic instances of this problem. Although numerous work exist on the static case of the RCPSP, research on the dynamic case is not that abundant. In addition, existing methods were not really developed for dynamic instances. We propose here to solve the dynamic RCPSP using a technique developed for solving dynamic constraint satisfaction problems: explanations. An explanation is a set of constraints (a subset of the current constraints system) that justifies an action made during the search process (deduction of new constraints, contradiction leading to a failure, etc.). An explanation can be considered as an explicit trace of constraint propagation. Explanations allow to efficiently undo past effects of a constraint and thus to dynamically add and remove constraints. We have also developed a branch and bound algorithm (inspired from the literature) that consist, in each node, in propagating the temporal and resources constraints (using core-times, tasks-intervals and resource-histogram techniques) while recording explanations. Moreover, we use a notion of distance (gap between the end of an activity and the beginning of another) which allows the modeling of temporal constraints within a unique concept. Our system is thus able to solve efficiently (i.e. without re-computing from scratch and in a short computation time) dynamic instances of the RCPSP (i.e. adding/removing precedence constraints, adding/removing activities and resources). Furthermore, our system, being very generic, allows the handling of several extensions of dynamic RCPSP (such as generalized precedence/disjunction/overlapping constraints, and variation of resource availability).
Document type :
Theses
Complete list of metadatas

Cited literature [85 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00008377
Contributor : Abdallah Elkhyari <>
Submitted on : Sunday, February 6, 2005 - 2:22:26 PM
Last modification on : Wednesday, November 29, 2017 - 10:10:33 AM
Long-term archiving on: : Friday, September 14, 2012 - 11:00:19 AM

Identifiers

  • HAL Id : tel-00008377, version 1

Citation

Abdallah Elkhyari. Outils d'aide à la décision pour des problèmes d'ordonnancement dynamiques. Génie logiciel [cs.SE]. Université de Nantes, 2003. Français. ⟨tel-00008377⟩

Share

Metrics

Record views

443

Files downloads

544