Nos travaux concernent la mise en œuvre de méthodologies pour la résolution et l'optimisation de la production en tenant compte des impératifs économiques, jouant aujourd'hui un rôle déterminant dans la conduite de la production industrielle. Pour le problème du job-shop flexible dans lequel les interactions entre les critères sont supposées disponibles, cinq critères ont été retenus : le Makespan, la charge critique, la charge totale, la pénalité de retards/avance et le coût de la production. Dans ce sens, nous avons, d'abord, traité le problème de décision et d'évaluation d'une solution et introduit ensuite trois approches intégrées, basées sur les algorithmes génétiques, améliorant les approches évolutionnistes existant dans la littérature : la méthode statique basée sur l'intégrale de Choquet, la méthode approchée basée sur le concept Paréto-optimalité ainsi que la méthode basée sur le concept de ε-dominance Paréto-optimalité. Les approches adoptées consistent à générer une variété de solutions optimales diversifiées dans l'espace de recherche de solutions, et d'aider le décideur, quand il ne peut pas donner une préférence particulière à l'une des fonctions objectif. Les résultats proposés, obtenus globalement pour l'ensemble des critères, ont été comparés, avec succès, avec ceux obtenus par d'autres approches existantes sur plusieurs benchmarks de complexités distinctes.

Nos travaux concernent l'étude d'une extension d'un problème d'ordonnancement bien connu sous l'appellation job shop. Nous appelons cette extension le General Flexible Job Shop Scheduling Problem (GFJSSP). Celui-ci se rencontre dans différents types d'ateliers ayant comme caractéristique commune d'être soumis à des contraintes dues à des ressources de transport. Le GFJSSP se caractérise par l'intégration de machines et robots flexibles. Le terme General induit par ailleurs la présence de robots dont la capacité est supposée unitaire dans notre étude, des temps opératoires bornés, et la possibilité de prise en compte d'emplacements de stockage spécifiques. Après avoir défini l'atelier et le problème correspondant à cette extension, nous avons proposé deux modélisations du GFJSSP ainsi défini : une première modélisation mathématique linéaire, et une modélisation graphique, qui correspond à une généralisation du graphe disjonctif couramment utilisé pour les problèmes de job shop. Nous avons ensuite abordé la résolution suivant deux étapes : tout d'abord en nous focalisant sur l'aspect séquencement des tâches de traitement et de transport, pour lequel nous avons élaboré deux méthodes heuristiques (de type Tabou et basée sur une procédure de shifting bottleneck améliorée) ; puis en intégrant dans un deuxième temps la problématique de l'affectation induite par la flexibilité de certaines ressources. Pour cette dernière étape, nous avons combiné les méthodes précédentes avec un algorithme génétique. L'algorithme hybride obtenu nous permet de résoudre des instances de la littérature correspondant à divers cas spécifiques, avec des résultats assez proches des meilleures méthodes dédiées. A termes, il pourrait être intégré dans un système d'aide à la décision général qui s'affranchirait de la phase d'identification préalable du type de job shop considéré, et serait adapté à la résolution de nombreux cas (avec ou sans problème d'affectation, temps de traitement fixes ou bornés, avec ou sans stockage, etc..).

Au cours de ces dernières années, les problèmes d'ordonnancement flexible ont largement attiré l'attention des chercheurs dans le domaine de la recherche opérationnelle. Ces problèmes présentent une difficulté supplémentaire du fait qu'une opération peut être exécutée par une ou plusieurs ressources devant être choisie(s) parmi d'autres candidates. L'objectif étant alors d'affecter et de séquencer les opérations sur les ressources en minimisant la durée d'exécution totale ou makespan. Dans cette étude, nous proposons de résoudre trois types de problèmes d'ordonnancement flexible : le flow shop hybride à plusieurs étages, à deux étages et le job shop flexible, en utilisant les méthodes arborescentes à base de divergences. Une étude expérimentale exhaustive a prouvé l'efficacité des différentes approches proposées pour les différents types de problèmes.

This study focused upon the manufacturing lead time management in California's electrical and electronic machinery, equipment, and supplies industry. Manufacturing firms with one hundred or more employees were invited to participate in the research. Six subproblems relating to manufacturing lead time were selected and six appropriate null hypotheses were tested. The subproblems identified (1) factors influencing manufacturing lead time, (2) production planning processes influencing manufacturing lead time accuracy, and (3) techniques reducing manufacturing lead time. These factors, production planning processes, and techniques were then IV investigated to determine the importance of each of them in two environments: (1) high-volume continuous-production (HVCP) and (2) job-shop production-to-order (JSPTO).

This thesis deals with optimization problems of planning and scheduling. There are using genetic algorithms which are inspired by evolution process. Main work is familiar with the problem of planning and scheduling, genetic algorithm and Petri nets. This knowledge was used to create applications that would with the use of genetic algorithms was able to solve planning problems and the resulting plans would be represented the Time Petri Net. In conclusion of the this thesis are presented obtained results and examples of field use.

In this thesis scheduling problems with transportation aspects are studied. Classical scheduling models for problems with multiple operations are the so-called shop-scheduling models. In these models jobs consisting of different operations have to be planned on certain machines in such a way that a given objective function is minimized. Each machine may process at most one operation at a time and operations belonging to the same job cannot be processed simultaneously. We generalize these classical shop-scheduling problems by assuming that the jobs additionally have to be transported between the machines. This transportation has to be done by robots which can handle at most one job at a time. Besides transportation times which occur for the jobs during their transport, also empty moving times are considered which arise when a robot moves empty from one machine to another. Two types of problems are distinguished: on the one hand, problems without transportation conflicts (i.e. each transportation can be performed without delay), and on the other hand, problems where transportation conflicts may arise due to a limited capacity of transport robots. In the first part of this thesis several new complexity results are derived for flow-shop problems with a single robot. Since very special cases of these problems are already NP-hard, in the second part of this thesis some techniques are developed for dealing with these hard problems in practice. We concentrate on the job-shop problem with a single robot and the makespan objective. At first we study the subproblem which arises for the robot when some scheduling decisions for the machines have already been made. The resulting single-machine problem can be regarded as a generalization of the traveling salesman problem with time windows where additionally minimal time-lags between certain jobs have to be respected and the makespan has to be minimized. For this single-machine problem we adapt immediate selection techniques used for other scheduling problems and calculate lower bounds based on linear programming and the technique of column generation. On the other hand, to determine upper bounds for the single-machine problem we develop an efficient local search algorithm which finds good solutions in reasonable time. This algorithm is integrated into a local search algorithm for the job-shop problem with a single robot. Finally, the proposed algorithms are tested on different test data and computational results are presented.

Decentralized decision-making is an active research topic in artificial intelligence. In a distributed system, a number of individually acting agents coexist. If they strive to accomplish a common goal, the establishment of coordinated cooperation between the agents is of utmost importance. With this in mind, our focus is on multi-agent reinforcement learning (RL) methods which allow for automatically acquiring cooperative policies based solely on a specification of the desired joint behavior of the whole system.The decentralization of the control and observation of the system among independent agents, however, has a significant impact on problem complexity. Therefore, we address the intricacy of learning and acting in multi-agent systems by two complementary approaches.First, we identify a subclass of general decentralized decision-making problems that features regularities in the way the agents interact with one another. We show that the complexity of optimally solving a problem instance from this class is provably lower than solving a general one.Although a lower complexity class may be entered by sticking to certain subclasses of general multi-agent problems, the computational complexitymay be still so high that optimally solving it is infeasible. Hence, our second goal is to develop techniques capable of quickly obtaining approximate solutions in the vicinity of the optimum. To this end, we will develop and utilize various model-free reinforcement learning approaches.Many real-world applications are well-suited to be formulated in terms of spatially or functionally distributed entities. Job-shop scheduling represents one such application. We are going to interpret job-shop scheduling problems as distributed sequential decision-making problems, to employ the multi-agent RL algorithms we propose for solving such problems, and to evaluate the performance of our learning approaches in the scope of various established scheduling benchmark problems.

No description available.

O ambiente de produção abordado neste trabalho é o Job Shop Flexível (JSF), uma generalização do Job Shop (JS). O problema de programação de tarefas, ou jobs, no ambiente JS é classificado por Garey; Johnson e Sethi (1976) como NP-Difícil e o JSF é, no mínimo, tão difícil quanto o JS. O JSF é composto por um conjunto de jobs, cada qual constituído por operações. Cada operação deve ser processada individualmente, sem interrupção, em uma única máquina de um subconjunto de máquinas habilitadas. O principal critério de desempenho considerado é a minimização dos atrasos dos jobs. São apresentados modelos de Programação Linear Inteira Mista (PLIM) para minimizar o atraso total e o instante de término da última operação, o makespan. São propostas novas regras de prioridade dos jobs, além de adaptações de regras da literatura. Tais regras são utilizadas por heurísticas construtivas e são aliadas a estratégias cujo objetivo é explorar características específicas do JSF. Visando aprimorar as soluções inicialmente obtidas, são propostas buscas locais e outros mecanismos de melhoria utilizados no desenvolvimento de três meta-heurísticas de diferentes categorias. Essas meta-heurísticas são: Iterated Local Search (ILS), classificada como meta-heurística de trajetória; Greedy Randomized Adaptive Search (GRASP), meta-heurística construtiva; e Artificial Bee Colony (ABC), meta-heurística populacional recentemente proposta. Esses métodos foram selecionados por alcançarem bons resultados para diversos problemas de otimização da literatura. São realizados experimentos computacionais com 600 instâncias do JSF, permitindo comparações entre os métodos de resolução. Os resultados mostram que explorar as características do problema permite que uma das regras de prioridade propostas supere a melhor regra da literatura em 81% das instâncias. As meta-heurísticas ILS, GRASP e ABC chegam a conseguir mais de 31% de melhoria sobre as soluções iniciais e a obter atrasos, em média, somente 2,24% superiores aos das soluções ótimas. Também são propostas modificações nas meta-heurísticas que permitem obter melhorias ainda mais expressivas sem aumento do tempo de execução. Adicionalmente é estudada uma versão do JSF com operações de Montagem e Desmontagem (JSFMD) e os experimentos realizados com um conjunto de 150 instâncias também indicam o bom desempenho dos métodos desenvolvidos. / The production environment addressed herein is the Flexible Job Shop (FJS), a generalization of the Job Shop (JS). In the JS environment, the jobs scheduling problem is classified by Garey; Johnson and Sethi (1976) as NP-Hard and the FJS is at least as difficult as the JS. FJS is composed of a set of jobs, each consisting of operations. Each operation must be processed individually, without interruption, in a single machine of a subset of enabled machines. The main performance criterion is minimizing the jobs tardiness. Mixed Integer Linear Programming (MILP) models are presented. These models minimize the total tardiness and the completion time of the last operation, makespan. New priority rules of jobs are proposed, as well as adaptations of rules from the literature. These rules are used by constructive heuristics and are combined with strategies aimed at exploiting specific characteristics of FSJ. In order to improve the solutions initially obtained, local searches and other improvement mechanisms are proposed and used in the development of metaheuristics of three different categories. These metaheuristics are: Iterated Local Search (ILS), classified as trajectory metaheuristic; Greedy Randomized Adaptive Search (GRASP), constructive metaheuristic, and Artificial Bee Colony (ABC), recently proposed population metaheuristic. These methods were selected owing to their good results for various optimization problems in the literature. Computational experiments using 600 FJS instances are carried out to allow comparisons between the resolution methods. The results show that exploiting the characteristics of the problem allows one of the proposed priority rules to exceed the best literature rule in about 81% of instances. Metaheuristics ILS, GRASP and ABC achieve more than 31% improvement over the initial solutions and obtain an average tardiness only 2.24% higher than the optimal solutions. Modifications in metaheuristics are proposed to obtain even more significant improvements without increased execution time. Additionally, a version called Disassembly and Assembly FSJ (DAFJS) is studied and the experiments performed with a set of 150 instances also indicate good performance of the methods developed.

Pour la résolution de problèmes d’ordonnancement d’ateliers de type flow-shop en industries pharmaceutiques et d’ateliers de type job-shop flexible, deux méthodes d’optimisation ont été développées : une méthode utilisant les algorithmes génétiques dotés d’un nouveau codage proposé et une méthode d’optimisation par essaim particulaire modifiée pour être exploitée dans le cas discret. Les critères retenus dans le cas de lignes de conditionnement considérées sont la minimisation des coûts de production ainsi que des coûts de non utilisation des machines pour les problèmes multi-objectifs relatifs aux industries pharmaceutiques et la minimisation du Makespan pour les problèmes mono-objectif des ateliers job-shop flexibles.Ces méthodes ont été appliquées à divers exemples d’ateliers de complexités distinctes pour illustrer leur mise en œuvre. L’étude comparative des résultats ainsi obtenus a montré que la méthode basée sur l’optimisation par essaim particulaire est plus efficace que celle des algorithmes génétiques, en termes de rapidité de la convergence et de l’approche de la solution optimale / For flexible job-shop and pharmaceutical flow-shop scheduling problems resolution, two optimization methods are considered: a genetic algorithm one using a new proposed coding and a particle swarm optimization one modified in order to be used in discrete cases.The criteria retained for the considered packaging lines in pharmaceutical industries multi-objective problems are production cost minimization and total stopping cost minimization. For the flexible job-shop scheduling problems treated, the criterion taken into account is Makespan minimization.These two methods have been applied to various work-shops with distinct complexities to show their efficiency.After comparison of these methods, the obtained results allowed us to notice the efficiency of the based particle swarm optimization method in terms of convergence and reaching optimal solution