O Problema de Escalonamento de Médicos (Physician Scheduling Problem) consiste em atribuir tarefas a médicos num horizonte de planejamento respeitando regras laborais, contratuais e de preferências pessoais de modo a satisfazer a demanda de serviços de um hospital. O problema lida majoritariamente com o objetivo de maximizar o atendimento dos requisitos de preferência pessoal, respeitando as restrições laborais e organizacionais. Sobre esta classe de problemas, vários métodos de resolução e suas variantes têm sido propostos na literatura. Ademais, mais características têm sido agregadas ao problema, tornando-o mais complexo e deste modo fazendo-se mais necessária a aplicação de métodos mais elaborados para a sua resolução. Neste trabalho são estudados, reformulados e propostos métodos de resolução baseados em programação matemática para tratar o problema de escalonamento acíclico de médicos em departamento de emergência de hospitais. O primeiro modelo tem como objetivo a minimização da soma ponderada dos desvios das restrições de distribuição. O segundo modelo tem como objetivo, a minimização do máximo dos desvios obtidos nas restrições de distribuição, a fim de se obter escalas mais equilibradas entre os médicos. Foram também propostas heurísticas baseadas na formulação matemática cujos resultados não foram competitivos com as dos modelos. Os modelos foram testados sobre um conjunto de instâncias fictícias resultantes de uma mescla entre instâncias benchmark e características do problema. Os resultados computacionais demonstram que formulação ponderada obteve solução ótima para grande parte das instâncias, embora os limitantes inferiores tenham sido majoritariamente fracos. Em relação ao segundo modelo, soluções ótimas não foram obtidas e os limitantes inferiores foram igualmente fracos. Relativamente a qualidade das escalas, o segundo modelo teve melhor comportamento comparando ao modelo de somas ponderadas. Dada a qualidade das soluções, nota-se a viabilidade da solução baseada em técnicas de otimização em detrimento da manual, pois esta ainda é mais suscetível de erros e acarreta um alto tempo para obtenção de solução. / The Physician Scheduling Problem consists in task assignment to physicians in a planning horizon considering a set of organizational rules, work regulations and individual preferences in order to satisfy an hospital wards work demand. The aim is to find a schedule which maximizes the satisfaction of individual preferences requirements while meeting work regulations and organizational rules. A plethora of solution methods and its variants have been proposed in the literature to solve this class of problem. Moreover, more features have been aggregated to the problem turning it into a more complex and thus estimulating the application of more elaborated methods to its decision. In this work we study, reshape and propose decision methods based in mathematical programming to handle non-ciclic physician scheduling problem in emergency wards. The first formulation targets the minimization of the weighted sum of distribution constraints deviations. The second formulation targets the minimization of the maximum deviations obtained at the distribution constraints aiming more balanced schedules between the physicians. Mathematical formulation heuristics were also proposed and the findings were not satisfactory as they were not competitive with the model. Experiments with our models were performed over a set of dummy instances, as result a of a mixture of benchmark instances and the considered problems features. From our experiments we have found that optimal solutions were obtained through the weighted sum model, despite the poor lower bounds. On the other hand, for the second model, no optimal solution was found and poor lower bounds were similarly obtained. Regarding to the schedules quality, the min-max model had a better performance comparing to the weighted sum model. Given the solutions quality we can assume that optimization based techniques are sustainable comparing to manual, because the latter is prone to errors and omissions and also critical in terms of solutions achievement time.

Existem vários problemas clássicos na área de pesquisa operacional que trabalham com o tema vinculado à designação de veículos em um sistema logístico, entre eles o Problema de Escalonamento de Veículos com Múltiplas Garagens (MDVSP). Esses modelos são largamente utilizados e representam uma das etapas essenciais para o planejamento de trânsito em massa (HAGHANI e BANIHASHEMI, 2002). Tratando-se de sistemas logísticos reais, dificilmente encontra-se um ambiente onde os veículos devem partir e chegar a uma única garagem, por isso torna-se necessário o planejamento das seqüências de viagens de modo a reduzir os custos de deslocamentos com o aproveitamento das múltiplas garagens distribuídas geograficamente. Infelizmente, considerando a complexidade exponencial do MDVSP, muitas vezes sua aplicação torna-se inviável na solução de problemas reais. Por essa razão, poucos trabalhos abordam o MDVSP de modo a conseguir solucionar o problema para uma grande quantidade de viagens e garagens. A maioria das pesquisas trabalha com instâncias inferiores a 500 viagens e quatro garagens, mostrando-se pouco aplicáveis. Esse estudo refere-se a um trabalho de pesquisa operacional que aborda soluções de problemas de escalonamento de veículos com múltiplas garagens (MDVSP) considerando sua aplicabilidade em sistemas reais. Tendo em vista a complexidade exponencial do MDVSP, nesse estudo optou-se por tratar o problema através de uma abordagem baseada na redução do espaço de estados e na utilização de heurísticas. Durante essa pesquisa três procedimentos de redução do espaço de estados foram adotados. Os resultados apontam que é possível reduzir em até 98% o número de variáveis nesses problemas sem comprometer uma solução satisfatória ou ótima. Além dos procedimentos de redução do espaço de estados, foi desenvolvido um procedimento de buscar a solução do MDVSP. Através desse último procedimento foi possível resolver o MDVSP com até 3000 viagens e oito garagens. Sendo assim, nesse estudo desenvolveram-se modelos que servem para o planejamento de um sistema logístico através da aplicação de cenários, com vistas a permitir a geração e análise de alternativas de escalonamento. Objetivou-se com isso, fornecer ao sistema logístico um modelo amplo que permita a escolha da ação mais conveniente e eficiente a ser tomada em modelos compostos por diversas garagens. / There are many classics problems in operations research concerning optimal assignment vehicles in logistical system. The multiple depot vehicle scheduling problem (MDVSP) is one of them. This problem is largely used to represent and solve mass transit planning (HAGHANI e BANIHASHEMI, 2002). Considering a real logistical system, it is very difficult to find out a situation where the vehicles must leave and come to only one depot. In general, the shipping company has several depots located at different sites in a network. In this way, it is strongly necessary to reduce cost through the planning of sequence trips taking into account multiple depots geographically distributed. Unfortunately, the exponential complexity of the MDVSP reduces, in the most cases, the applicability of this problem in the real world. For this reason, few researchers address the MDVSP to solve real world problems considering a large number of trips and depots. The majority of the research dealing with the MDVSP works with instances lower than 500 trips and four depots, what can be considered a major constraint for its practical use. The main objective of this work is to solve the MDVSP for very large instances. A state space reduction approach combined with heuristic procedures are developed to obtain a realistic way of solving this complex problem. In this research, three state space reduction procedures were developed. The results appointed that is possible to reduce until 98% of variables in the MDVSP without jeopardizing an optimal solution. Furthermore, heuristic procedures were developed to obtain solutions without relaxing any realworld constraint of the problem. The solution procedure developed was compared with wellknown available instances. The method is able to solve the MDVSP with 3000 trips and eight depots in less than 11 minutes. Although the solution process does not obtain the best solution in all tested instances, it is by far the quickest.

Titel: Planeringsmetoder i processindustrin: En fallstudie för AAK AB Författare: Mirlinda Brahimi och William JonassonHandledare: Peter BerlingExaminator: Helena Forslund  Bakgrund: I dagsläget har AAK svårigheter att finna tillräcklig kapacitet för att kunna möta sina kunders ordrar i vissa fall. Det förekommer att de full belastar sin produktionskapacitet och då de måste tacka nej till vissa kunders beställningsordrar. Studien grundar sig i att undersöka och skapa mer förståelse för cyklisk planering eller schemaläggning, då det finns ett behov av att finna sätt att samproducera och gruppera produkter i planeringen. Olika metoder granskas för att upptäcka en lämplig eller förbättrad planeringsmetod som kan skapa produktionsfokus i cykler som vidare bidrar till att minska kapacitetsproblemen och öka produktionseffektiviteten.  Syfte: Studiens syfte är att kartlägga och fördjupa sig i planeringsmiljön på AAK för att förbättra deras produktionsplanering i verksamheten. Det görs främst genom att förenkla och placera AAK:s planeringsmiljö i vilken miljötyp företaget tillhör samt dess förutsättningar som förekommer i miljön. Utifrån miljötypen så identifieras olika planeringsmetoder för att ta reda om dess tillämpbarhet kan realiseras på praktiken och faktiskt stabilisera planeringsmiljöns komplexitet på företaget såväl som det ska leda till en ökad fabriksproduktion.Metod: Studiens metod består av en kvalitativ studie, men också en kvantitativ då det görs beräkningar av olika modeller för det dataunderlag som erhållits för AAK:s kartongtappning. Den teoretiska referensramen skapades främst genom sökning av litterära källor och vetenskapliga artiklar. Studiens empiriska data har insamlats genom ostrukturerade- och semistrukturerade intervjuer av olika personer i studiens fallföretag. Resultat och slutsats: Studiens resultat visar att AAK:s miljötyp liknar en repetitiv masstillverkning. Powers-of-two beräkningar visade på goda resultat för kartongtappningen, men skulle kunna anpassas utifrån ett Product Wheel för att ta vara på båda metodernas fördelar. Mycket tyder på liknande resultat för andra processindustrier, då miljötypen är vanlig inom processindustrin.

[pt] Esta dissertação de mestrado analisa um problema de programação de máquinas em série com datas de entrega e ganhos cumulativos sob incerteza. Em particular, este trabalho considera situações reais na quais os tempos de processamento e datas de liberação são estocásticos. O objetivo principal deste trabalho é a resolução deste problema de programação de máquinas em série em um ambiente estocástico buscando analisar a relação entre diferentes niveis de incerteza e o benefício esperado. Visando atingir este objetivo, primeiramente uma heurística é proposta utilizando-se da técnica de biased-randomization para a versão determinística do problema. Então, esta heurística é extendida para uma metaheurística a partir do encapsulamento dentro da estrutura de um variable neighborhood descend. Finalmente, a metaheurística é extendida para uma simheurística a partir da incorporação da simulação de Monte Carlo. De acordo com os experimentos computacionais, o nível de incerteza tem um impacto direto nas soluções geradas pela simheurística. Além disso, análise de risco foram desenvolvidas utilizando as conhecidas métricas de risco: value at risk e conditional value at risk. / [en] This master s thesis analyzes the Permutation Flow-shop Scheduling Problem with Delivery Dates and Cumulative Payoffs under uncertainty conditions. In particular, the work considers the realistic situation in which processing times and release dates are stochastics. The main goal is to solve this Permutation Flow-shop problem in the stochastic environment and analyze the relationship between different levels of uncertainty and the expected payoff. In order to achieve this goal, first a biased-randomized heuristic is proposed for the deterministic version of the problem. Then, this heuristic is extended into a metaheuristic by encapsulating it into a variable neighborhood descent framework. Finally, the metaheuristic is extended into a simheuristic by incorporating Monte Carlo simulation. According to the computational experiments, the level of uncertainty has a direct impact on the solutions provided by the simheuristic. Moreover, a risk analysis is performed using two well-known metrics: the value at risk and the conditional value at risk.

Olofsfors is a steel product manufacturer in Nordmaling, Sweden, producing steel edges for snowplows, tracks for forest machines, and wear edges for buckets on heavy equipment. Most of their products are heated to 900◦ C and then cooled down in water, so-called quenching, during the hardening process. A group of ovens and quench machines together form an oven system and this is used for the hardening. Since it takes a long time for the ovens to reach operating temperature, they are always kept on, which is why it is important to utilize them as effectively as possible. This project investigates the potential utilization increase of one of the three oven systems in the wear edge production unit. This oven system is part of a production line that consists of a saw and a mill, and can process products up to two meters in length, and is hereon called the two-meter line. The two-meter line has a natural inflow through the saw, but raw material produced in other parts of the factory can also be fetched from another inlet. The use of the other inlet is limited by the operator of the two-meter line who has to fetch the material with a forklift. This could be automated so that the operator would not have to handle this inlet. The purpose is to investigate the potential increases in utilization of the oven system for different degrees of automation in order to make the most of the machines and the operator at the two-meter line. In the end, a recommendation is given with a set of ideal properties of the investment that could improve productivity the most. The main method applied in order to explore the potential use of the oven system is a re-entrant flow shop scheduling model. As preceding steps, the production line is first mapped in order to find potential routes for different product families, then the order quantities in the production data are translated into jobs to be scheduled with the help of packing problems and batching rules. The scheduling model of the production line is then solved heuristically with a genetic algorithm based on the sequence of jobs entering the production line followed by a method for creating a deterministic schedule based on this initial sequence of jobs. Lastly, a sensitivity analysis is applied to the processing time for the steps performed by the operator to evaluate the results' robustness. The conclusion is that there is a substantial potential to increase the utilization of the oven system of the two-meter line. The largest potential is when the operator is not actively working at the production line; a maximum of 15.6 h on average. There does also exist a potential to increase utilization while the operator is working at the production line; a maximum of 3.9 h on average. The automation degree needed is high in both cases but due to different reasons. When the operator is not working, the automatic solution needs to work without supervision for longer periods of time, while, in the other case, it needs to be smart enough to adjust to not disturb the operator’s work. For the future, the recommendation is to focus the next step on finding investment options that could exploit the time when the operator is not working. By further specifying the potential investment alternatives, the cost factor can be added to the analysis as well. / Olofsfors AB är en stålproduktstillverkare i Nordmaling, Sverige, som producerar vägstål till bland annat snöplogar, band till skogsmaskiner och slitstål till entreprenadmaskiner. De flesta av deras produkter hettas upp till 900 C och släcks sedan i vatten under härdningsprocessen. En grupp av ugnar och härdmaskiner kallas tillsammans för ett ungsystem och det används till härdningen. Eftersom det tar lång tid att värma upp ugnarna står de alltid på-slagna och det är därför viktigt att använda dem så effektivt som möjligt.  I detta projekt har potentialen att öka användandet av ett av tre ugnsystem i slitstålsproduktionen undersökts.  Ugnsystemet i fråga är en del av en produktionslinje som också består av en såg och en fräs och kan härda artiklar med längder upp till två meter och kallas därför här tvåmeterslinjen. Den naturliga ingången för råmaterial i produktionslinjen är genom sågen, men det finns även en alternativ ingång för råmaterial som förbehandlats i tidigare produktionssteg i fabriken. Användandet av den andra ingången till produktionlinjen begränsas av att operatören i produktionslinjen måste hämta materialet med truck. Detta in-flöde skulle gå att automatisera så att operatören inte skulle behöva hämta dessa artiklar.  Syftet är att undersöka det potentiella ökade nyttjandet av ugnsystemet för olika grader av automation för att bäst använda maskiner och operatör i tvåmeterslinjen. I slutet ges en rekommendation gällande vilka egenskaper investeringen bör ha för att öka produktiviteten mest.  Huvudmetoden för att undersöka möjligt ökat nyttjande av ugnarna är en schemaläggningsmodel. Som underliggande steg kartläggs först produktionslinjen och de olika rutter som olika produktfamiljer tar genom produktionslinjen. Produktkvantiteterna för varje order i produktionsdatan omvandlas sedan till jobb som kan schemaläggas genom packningsproblem och regler för laststorlekar i de olika maskinerna. Schemaläggningsmodellen löses sedan heuristiskt med hjälp av en genetisk algoritm som bestämmer den initiala sekvensen av jobben i första steget, tillsammans med en deterministisk metod för att skapa ett helt schema baserat på den initiala sekvensen av jobben. Slutligen genomförs en känslighetsanalys på processtiderna för steg som motsvarar operatören för att undersöka hur robust resultatet är.    Slutsatsen är att det finns en stor potential att öka nyttjandet av ugnsystemet i tvåmeterslinjen. Den största potentialen är när operatören inte arbetar aktivt vid produktionslinjen, med ett maximum på ca 15,6 h per dag. Det finns också en möjlighet att utöka nyttjandet av ugnarna under tiden som operatören arbetar aktivt med ordrar och outnyttjad tid då är 3,9 h i genomsnitt. Graden av automation är hög oberoende av vilken tid som ska utnyttjas men på grund av olika anledning. Om tiden då operatören inte aktivt jobbar utnyttjas, måste den automatiserade lösningen fungera autonomt under längre tid. Om den istället förväntas fungera parallellt med operatören måste den anpassas smart så att den inte stör operatörens arbete och flöde. Rekommendationen är att fokusera på att hitta konkreta investeringsalternativ som utnyttjar tiden då operatören inte aktivt arbetar för att få bättre kostnadsunderlag att ha med i den vidare analysen.

Scheduling problem and shift design problems are well known NP-hard problems within the optimization area. Often time, the two problems are studied individually. In this thesis however, we are looking at the combination of both problems. More specifically, the aim of this thesis is to suggest an optimal scheduling policy given that there are no predefined shifts to begin with. The duration of a shift, along with the start and end time may vary. Thus we have proposed to split the problem into two sub-problems: weekly scheduling problem and daily scheduling problem. As there are no exact solution methods that are feasible, two meta-heuristics method has been employed to solve the sub-problems: Simulated Annealing (SA) and Genetic Algorithm (GA). We have provided proofs of concepts for both methods as well as explored the scalability. This is especially important as the number of employee is expected to grow significantly throughout the year. The results obtained has shown to be promising and can be built upon for further capabilities. / Schemaläggning och skiftdesignsproblem är välkända och välstuderade NP-svåra beslutsproblem inom optimeringsområdet. Oftast så studeras dessa problem enskilt, men i detta arbete så studeras en kombination av båda problemen. Mer specifikt är målet med detta arbete att föreslå ett förnuftigt handlingsätt till att skapa ett veckoschema där skift inte är predefinierade för alla veckor. Starttiden, sluttiden och varaktigheten av ett skift kan förändras från vecka till vecka. Därför har problemet delats upp till två delar: Veckoschemaläggnings- och dagsschemaläggningsproblem. Trots uppdelningen så är båda delproblem för komplexa för att lösas exakt. Därför har två metaheuristiska metoder använts som lösningsmetoder: Simulerad Glödgning och Genetisk Algoritm. I detta arbete bevisas båda lösningsmetoderna till att vara bra nog, och dessutom studeras även skalbarheten av modellen. Detta senare är särskilt viktigt eftersom antal anställda som ska schemaläggas förväntas att öka genomåren. De erhållna resultaten har visat sig vara lovande och bevisligen så kan modellen expanderas med er villkor

We develop a detailed, large-scale, airspace planning and collaborative decision-making model (APCDM), that is part of an $11.5B, 10-year, Federal Aviation Administration (FAA)-sponsored effort to increase U.S. National Airspace (NAS) capacity by 30 percent. Given a set of flights that must be scheduled during some planning horizon, we use a mixed-integer programming formulation to select a set of flight plans from among alternatives subject to flight safety, air traffic control workload, and airline equity constraints. Novel contributions of this research include three-dimensional probabilistic conflict analyses, the derivation of valid inequalities to tighten the conflict safety representation constraints, the development of workload metrics based on average (and its variance from) peak load measures, and the consideration of equity among airline carriers in absorbing the costs related to re-routing, delays, and cancellations. We also propose an improved set of flight plan cost factors for representing system costs and investigating fairness issues by addressing flight dependencies occurring in hubbed operations, as well as market factors such as schedule convenience, reliability, and the timeliness of connections. The APCDM model has potential use for both tactical and strategic applications, such as air traffic control in response to severe weather phenomenon or spacecraft launches, FAA policy evaluation, Homeland Defense contingency planning, and military air campaign planning. The model is tested to consider various airspace restriction scenarios imposed by dynamic severe weather systems and space launch Special Use Airspace (SUA) impositions. The results from this model can also serve to augment the FAA's National Playbook of standardized flight profiles in different disruption-prone regions of the National Airspace. / Ph. D.

Résumé indisponible. / In this thesis we have concerned ourselves with university timetabling problems both course timetabling and examination timetabling problems. Most of the timetabling problems are computationally NP-complete problems, which means that the amount of computation required to find solutions increases exponentially with problem size. These are idiosyncratic nature problems, for example different universities have their own set of constraints, their own definition of good timetable, feasible timetable and their own choice about the use of constraint type (as a soft or hard constraint). Unfortunately, it is often the case that a problem solving approach which is successfully applied for one specific problem may not become suitable for others. This is a motivation, we propose a generalized problem which covers many constraints used in different universities or never used in literature. Many university timetabling problems are sub problems of this generalized problem. Our proposed algorithms can solve these sub problems easily, moreover constraints can be used according to the desire of user easily because these constraints can be used as reference to penalty attached with them as well. It means that give more penalty value to hard constraints than soft constraint. Thus more penalty value constraints are dealt as a hard constraint by algorithm. Our algorithms can also solve a problem in two phases with little modification, where in first phase hard constraints are solved. In this work we have preferred and used two phase technique to solve timetabling problems because by using this approach algorithms have broader search space in first phase to satisfy hard constraints while not considering soft constraints at all. Two types of algorithms are used in literature to solve university timetabling problem, exact algorithms and approximation algorithms. Exact algorithms are able to find optimal solution, however in university timetabling problems exact algorithms constitute brute-force style procedures. And because these problems have the exponential growth rates of the search spaces, thus these kinds of algorithms can be applied for small size problems. On the other side, approximation algorithms may construct optimal solution or not but they can produce good practically useable solutions. Thus due to these factors we have proposed approximation algorithms to solve university timetabling problem. We have proposed metaheuristic based techniques to solve timetabling problem, thus we have mostly discussed metaheuristic based algorithms such as evolutionary algorithms, simulated annealing, tabu search, ant colony optimization and honey bee algorithms. These algorithms have been used to solve many other combinatorial optimization problems other than timetabling problem by modifying a general purpose algorithmic framework. We also have presented a bibliography of linear integer programming techniques used to solve timetabling problem because we have formulated linear integer programming formulations for our course and examination timetabling problems. We have proposed two stage algorithms where hard constraints are satisfied in first phase and soft constraints in second phase. The main purpose to use this two stage technique is that in first phase hard constraints satisfaction can use more relax search space because in first phase it does not consider soft constraints. In second phase it tries to satisfy soft constraints when maintaining hard constraints satisfaction which are already done in first phase. (...)

Le domaine de l'embarqué connaît depuis quelques années un essor important avec le développement d'applications de plus en plus exigeantes en calcul auxquels les architectures traditionnelles à base de processeurs (mono/multi cœur) ne peuvent pas toujours répondre en termes de performances. Si les architectures multiprocesseurs ou multi cœurs sont aujourd'hui généralisées, il est souvent nécessaire de leur adjoindre des circuits de traitement dédiés, reposant notamment sur des circuits reconfigurables, permettant de répondre à des besoins spécifiques et à des contraintes fortes particulièrement lorsqu'un traitement temps-réel est requis. Ce travail présente l'étude des problèmes d'ordonnancement dans les architectures hétérogènes reconfigurables basées sur des processeurs généraux (CPUs) et des circuits programmables (FPGAs). L'objectif principal est d'exécuter une application présentée sous la forme d'un graphe de précédence sur une architecture hétérogène CPU/FPGA, afin de minimiser le critère de temps d'exécution total ou makespan (Cmax). Dans cette thèse, nous avons considéré deux cas d'étude : un cas d'ordonnancement qui tient compte des délais d'intercommunication entre les unités de calcul CPU et FPGA, pouvant exécuter une seule tâche à la fois, et un autre cas prenant en compte le parallélisme dans le FPGA, qui peut exécuter plusieurs tâches en parallèle tout en respectant la contrainte surfacique. Dans un premier temps, pour le premier cas d'étude, nous proposons deux nouvelles approches d'optimisation, GAA (Genetic Algorithm Approach) et MGAA (Modified Genetic Algorithm Approach), basées sur des algorithmes génétiques. Nous proposons également de tester un algorithme par séparation et évaluation (méthode Branch & Bound). Les approches GAA et MGAA proposées offrent un très bon compromis entre la qualité des solutions obtenues (critère d'optimisation de makespan) et le temps de calcul nécessaire à leur obtention pour résoudre des problèmes à grande échelle, en comparant à la méthode par séparation et évaluation (Branch & Bound) proposée et l'autre méthode exacte proposée dans la littérature. Dans un second temps, pour le second cas d'étude, nous avons proposé et implémenté une méthode basée sur les algorithmes génétiques pour résoudre le problème du partitionnement temporel dans un circuit FPGA en utilisant la reconfiguration dynamique. Cette méthode fournit de bonnes solutions avec des temps de calcul raisonnables. Nous avons ensuite amélioré notre précédente approche MGAA afin d'obtenir une nouvelle approche intitulée MGA (Multithreaded Genetic Algorithm), permettent d'apporter des solutions au problème de partitionnement. De plus, nous avons également proposé un algorithme basé sur le recuit simulé, appelé MSA (Multithreaded Simulated Annealing). Ces deux approches proposées, basées sur les méthodes métaheuristiques, permettent de fournir des solutions approchées dans un intervalle de temps très raisonnable aux problèmes d'ordonnancement et de partitionnement sur système de calcul hétérogène / The domain of the embedded systems becomes more and more attractive in recent years with the development of increasing computationally demanding applications to which the traditional processor-based architectures (either single or multi-core) cannot always respond in terms of performance. While multiprocessor or multicore architectures have now become generalized, it is often necessary to add to them dedicated processing circuits, based in particular on reconfigurable circuits, to meet specific needs and strong constraints, especially when real-time processing is required. This work presents the study of scheduling problems into the reconfigurable heterogeneous architectures based on general processors (CPUs) and programmable circuits (FPGAs). The main objective is to run an application presented in the form of a Data Flow Graph (DFG) on a heterogeneous CPU/FPGA architecture in order to minimize the total running time or makespan criterion (Cmax). In this thesis, we have considered two case studies: a scheduling case taking into account the intercommunication delays and where the FPGA device can perform a single task at a time, and another case taking into account parallelism in the FPGA, which can perform several tasks in parallel while respecting the constraint surface. First, in the first case, we propose two new optimization approaches GAA (Genetic Algorithm Approach) and MGAA (Modified Genetic Algorithm Approach) based on genetic algorithms. We also propose to compare these algorithms to a Branch & Bound method. The proposed approaches (GAA and MGAA) offer a very good compromise between the quality of the solutions obtained (optimization makespan criterion) and the computational time required to perform large-scale problems, unlike to the proposed Branch & Bound and the other exact methods found in the literature. Second, we first implemented an updated method based on genetic algorithms to solve the temporal partitioning problem in an FPGA circuit using dynamic reconfiguration. This method provides good solutions in a reasonable running time. Then, we improved our previous MGAA approach to obtain a new approach called MGA (Multithreaded Genetic Algorithm), which allows us to provide solutions to the partitioning problem. In addition, we have also proposed an algorithm based on simulated annealing, called MSA (Multithreaded Simulated Annealing). These two proposed approaches which are based on metaheuristic methods provide approximate solutions within a reasonable time period to the scheduling and partitioning problems on a heterogeneous computing system

Este trabalho aborda o problema multiestágio de planejamento e programação da produção em indústrias cervejeiras. O processo de fabricação de cerveja pode ser dividido em duas etapas principais: preparação do líquido e envase. A primeira etapa ocorre, na maior parte do tempo, dentro de tanques de fermentação e maturação. A segunda ocorre nas linhas de envase, podendo ter início assim que o líquido estiver pronto nos tanques. O tempo de preparação do líquido demora vários dias, enquanto que na maioria das indústrias de bebidas carbonatadas este tempo é de no máximo algumas horas. O objetivo deste estudo é obter planos de produção viáveis que visam otimizar as decisões de programação envolvidas nestes processos. Visitas a cervejarias no Brasil e em Portugal foram realizadas para uma maior familiaridade do processo de produção e dados foram coletados. Modelos de programação inteira mista para representar o problema foram desenvolvidos, baseados em abordagens CSLP (The Continuous Setup Lot-Sizing Problem), GLSP (General Lot Sizing and Scheduling Problem), SPL (Simple Plant Location Problem) e ATSP (Asymmetric Travelling Salesman Problem). Os resultados mostram que os modelos são coerentes e representam adequadamente o problema, entretanto, mostram-se difíceis de serem resolvidos na otimalidade. Esta dificuldade de resolução dos modelos motivou o desenvolvimento de procedimentos MIP-heurísticos, como também de uma metaheurística GRASP (Greedy Randomized Adaptive Search Procedure). As soluções obtidas pelos procedimentos heurísticos são de boa qualidade, quando comparadas ao melhor limitante inferior encontrado por meio da resolução dos modelos matemáticos. Os testes computacionais foram realizados utilizando instâncias geradas com base em dados reais. / This study deals with the multistage lot-sizing and scheduling problem in breweries. The brewing process can be divided into two main stages: preparation and filling of the liquid. The first stage occurs most of the time in fermentation and maturation tanks. The second stage occurs in the filling lines and it can start as soon as the liquid gets ready. The preparation time of the liquid takes several days, while in the carbonated beverage industries this time is at most a few hours. The purpose of this study is to obtain feasible production plans aimed at optimizing the decisions involved in these processes. Visits to brewery industries in Brazil and Portugal were held to a greater familiarity of the production process and data were collected. Mixed integer programming models have been developed to represent the problem, based on approaches for the CSLP (The Continuous Setup Lot-Sizing Problem), GLSP (General Lot Sizing and Scheduling Problem), SPL (Simple Plant Location Problem) and ATSP (Asymmetric Travelling Salesman Problem). The results show that the models are consistent and adequately represent the problem; however, they are difficult to be solved at optimality. This motivated the development of MIP-heuristic procedures, as well as a meta-heuristic GRASP (Greedy Randomized Adaptive Search Procedure). The obtained solutions by the heuristics are of good quality, when compared to the best lower bound found by solving the mathematical models. The tests were conducted using generated instances based on real data.