Made available in DSpace on 2016-03-15T19:37:43Z (GMT). No. of bitstreams: 1 Luiz Antonio Carraro.pdf: 1226702 bytes, checksum: 3cef29694a4e26f233b0aae16da69cf0 (MD5) Previous issue date: 2012-02-16 / Universidade Presbiteriana Mackenzie / A container is a broadly used solution for the cargo storage to be transported between ports, playing a central role in international trade. Consequently, ships grew in size in order to maximize their container transportation capacity in each trip. Due to increasing demand, container terminals face the challenges of increasing their service capacity and optimizing the loading and unloading time of ships. Optimization problems, such as these, often present features that make it impossible to obtain closed analytical solutions, requiring iterative search procedures in high-dimensional spaces, or subject to a combinatorial explosion of possible solutions. This dissertation presents the proposal of a novel meta-heuristic based on the Clonal Selection Algorithm, named MRC, to minimize the number of reshuffles in operations involving piles of containers. The performance of the proposed model was evaluated through simulations and results comparison with those obtained by algorithms from the literature under the same test conditions. The results obtained show that MRC is competitive in terms of minimizing the need of reshuffles, besides presenting a reduced processing time compared with models of similar performance. / A utilização de contêineres é uma solução amplamente adotada para o armazenamento da carga a ser transportada entre portos, tornando-se de grande importância no comércio internacional e, consequentemente, navios cresceram de tamanho com o objetivo de transportar a maior quantidade possível de contêineres em cada viagem. Devido à crescente demanda, terminais de contêineres enfrentam os desafios de aumentar a sua capacidade de atendimento e otimizar os tempos de carregamento e descarregamento de navios. Problemas de otimização como estes geralmente apresentam características que inviabilizam a obtenção de soluções analíticas fechadas, requerendo processos iterativos de busca em espaços de dimensão muitas vezes elevada, ou ainda sujeitos a explosão combinatória de possíveis soluções. Esta dissertação apresenta a proposta de uma meta-heurística bioinspirada baseada no Algoritmo de Seleção Clonal para a minimização de rearranjos em operações que envolvem pilhas de contêineres, denominado MRC. O desempenho do algoritmo foi avaliado por meio de simulações e comparação dos resultados com os obtidos por algoritmos da literatura sob as mesmas condições de teste. Os resultados obtidos permitem concluir que o MRC possui resultados competitivos em termos de minimização de rearranjos, além de apresentar um tempo de processamento reduzido quando comparado aos modelos tradicionalmente empregados na solução desse tipo de problema.

Made available in DSpace on 2016-08-17T14:53:26Z (GMT). No. of bitstreams: 1 Dissertacao Aleksandro Costa.pdf: 1706153 bytes, checksum: 8d61027896dbab484303f78ed17b9b70 (MD5) Previous issue date: 2014-02-28 / FUNDAÇÃO DE AMPARO À PESQUISA E AO DESENVOLVIMENTO CIENTIFICO E TECNOLÓGICO DO MARANHÃO / This work has the objective to develop an analytical model that makes prediction of the behavior of the algorithm as a function of the design parameters (step adaptation, kernel function and its parameters).We use a non-quadratic function based on kernel, performing a nonlinear transformation of the input space filtering applied on line. Was developed and implemented in the system for adaptive filtering based on Kernel, which provides an analysis of the behavior of KRLS algorithm as well as its properties of convergence. It applies a kernel function in the cost function from the non-recursive quadratic function of an even power, which minimizes the error, defined as the expectation of the cumulative cost of actions taken along a sequence of steps. It appears that this approach allows the determination of the parameters of the problem with greater reliability and robustness and lower cost compared with traditional algorithms (RLS, KRLS, RNQ) . / Este trabalho tem como objetivo desenvolver um modelo analítico que faça a previsão do comportamento do algoritmo RLS como uma função dos parâmetros de projeto (passo de adaptação, função kernel e seus parâmetros). Utiliza-se uma função não quadrática baseado em kernel, realizando uma transformação não linear do espaço de entrada aplicada à filtragem. Foi desenvolvido e implementado na redução de ruídos para a filtragem adaptativa baseada em Kernel, que fornece uma análise do comportamento do algoritmo KRLS, bem como de suas propriedades de convergência. Aplica-se uma função kernel na função de custo a partir da função recursiva não quadrática de quarta potência, que minimiza o erro, definido como a expectativa do custo cumulativo de ações tomadas ao longo de uma sequência de passos. Verifica-se que essa abordagem possibilita a determinação dos parâmetros do problema com uma maior confiabilidade e robustez e o menor custo, quando comparado com algoritmos tradicionais (RLS, KRLS, RNQ).

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-04-17T20:10:32Z No. of bitstreams: 1 danielrodriguespereira.pdf: 736772 bytes, checksum: d15b2f08bb14ed58ae985f6123258ed5 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-04-18T13:51:41Z (GMT) No. of bitstreams: 1 danielrodriguespereira.pdf: 736772 bytes, checksum: d15b2f08bb14ed58ae985f6123258ed5 (MD5) / Made available in DSpace on 2017-04-18T13:51:41Z (GMT). No. of bitstreams: 1 danielrodriguespereira.pdf: 736772 bytes, checksum: d15b2f08bb14ed58ae985f6123258ed5 (MD5) Previous issue date: 2017-02-07 / Apresentamos neste trabalho o algoritmo de pontos interiores e direções viáveis denominado FDIPA para resolução de problemas de otimização definido por uma função diferenciável e por restrições de desigualdades. O algoritmo gera uma sequência de pontos interiores a partir de um dado ponto inicial também de interior e converge globalmente com ordem superlinear para um par Karush-Kuhn-Tucker do problema. A cada iteração uma direção de descida da função potencial é calculada inicialmente pela resolução de um sistema nas variáveis dual e primal. Apresentamos também o algoritmo FDA para resolução de problemas de complementaridade definido por uma função diferenciável e não linear. Mostramos a equivalência entre os dois métodos no sentido de gerarem as mesmas direções de descida, viável e de restauração a partir de uma atualização dos multiplicadores de Lagrange do problema de otimização. Realizamos uma comparação entre os métodos em uma coletânea de problemas de complementaridade. / In this work we present the algorithm of internal points and viable directions denominated FDIPA to solve optimization problems defined by a differentiable function and by inequalities restrictions. The algorithm generates a sequence of interior points from a given interior starting point and converges globally with superlinear order to a Karush-Kuhn-Tucker pair of the problem. At each iteration a descent direction of the potential function is calculated initially by the solution of a system in the dual and primal variables. We also present the FDA algorithm to solve complementarity problems defined by a non-linear differentiable function. We show the equivalence between the two methods in the sense that they generate the same descent, feasible and restoring directions from an update to the Lagrange multipliers of the optimization problem. We perform a comparison between the two methods in a collection of complementarity problems.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-04-25T15:55:49Z No. of bitstreams: 1 larafurtadobastos.pdf: 4369558 bytes, checksum: 7b36e77b964a5ec919c2c9967a654a03 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-04-26T12:06:02Z (GMT) No. of bitstreams: 1 larafurtadobastos.pdf: 4369558 bytes, checksum: 7b36e77b964a5ec919c2c9967a654a03 (MD5) / Made available in DSpace on 2017-04-26T12:06:02Z (GMT). No. of bitstreams: 1 larafurtadobastos.pdf: 4369558 bytes, checksum: 7b36e77b964a5ec919c2c9967a654a03 (MD5) Previous issue date: 2016-09-08 / O presente trabalho apresenta a adaptação e utilização de um algoritmo da área de inteligência artificial evolucionária, bioinspirado no sistema de ecolocalização de morcegos, para resolver o problema da localização global de robôs móveis em ambientes bidimensionais com mapas conhecidos. Sabe-se, por meio da literatura, que a localização de robôs baseada apenas em dedução via hodometria, do inglês deduced reckoning ou dead-reckoning, acumula diversos erros de origem estocástica, os quais não podem ser eliminados de maneira determinística, fazendo-se necessários métodos de filtragem estatística para a correta obtenção da localização. Dentre as diversas alternativas conhecidas para solucionar o problema de localização, escolheu-se o Método Recursivo de Monte Carlo, também denominado por Filtro de Partículas, para comparação com os resultados obtidos pelo algoritmo de morcego, por suas características multimodais e não-paramétricas, sendo este um algoritmo clássico na área de localização robótica. O algoritmo de morcegos, do inglês Bat Algorithm, é um método recursivo de otimização de estados de um sistema que se encontra num ambiente multimodal. É bioinspirado nos sistemas de ecolocalização encontradas em morcegos e outros animais na natureza. Nos resultados de comparação entre ambos os métodos, a técnica proposta demonstrou melhores resultados tanto para o erro entre a localização real e a estimada pelos métodos quanto para o número de iterações necessárias para alcançar a solução e, consequentemente, o tempo de convergência do algoritmo. Para o desenvolvimento deste trabalho, utilizou-se o programa Matlab R integrado com a plataforma ROS, juntamente com o robô móvel terrestre Pioneer P3-DX para os resultados simulados e reais. / This work presents the adaptation and use an algorithm from evolutionary artificial intelligence area, bioinspired in the echolocation system of bats to solve the problem of global location for mobile robots in two-dimensional environments with known maps. It is widely known in literature that the localization of robots based only on deduced reckoning accumulates many stochastic errors, which cannot be eliminated deterministically, requesting statistical filtering methods to obtain the correct location. Among the various alternatives known to solve the problem of localization, we chose the Recursive Method of Monte Carlo, also kown as Particle Filter, for comparison purposes with the results obtained by the Bat Algorithm, because of its multimodal and nonparametric features, and alse because it is a classic algorithm in robotics localization area. The Bat Algorithm is a recursive optimization method of system states immerse in multimodal environments. It is bioinspired in the echolocation systems found in bats and other animals in nature. In comparison results between the two methods, the proposed technique showed the best results for both localization error and the number of iterations required to reach the solution, and consequently the algorithm convergence time. To develop this work, the Matlab software was used with the ROS framework along with the terrestrial mobile robot Pioneer P3-DX for simulated and real results.

Orientadores: Akebo Yamakami, Tatiane Regina Bonfim / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-19T06:06:35Z (GMT). No. of bitstreams: 1 Carvalho_MarciaBragade_D.pdf: 1895321 bytes, checksum: ae40a5cf6d05e99795952c1a7c6bed79 (MD5) Previous issue date: 2011 / Resumo: Este trabalho propõe novas abordagens híbridas baseadas em técnicas da computação bio-inspirada para o problema de escalonamento do tipo Job Shop. Como o problema do tipo job shop pertence a classe NP-difícil e não existe algoritmo exato capaz de solucionar todos os tipos deste problema. Normalmente é necessária a elaboração de métodos de resolução mais sofisticados para contornar essa alta complexidade. Desta forma, nesta tese propomos abordagens híbridas baseadas em algoritmo memético e algoritmo de otimização por colônia de formigas a fim de contornar essa complexidade e ser capaz de explorar eficientemente o espaço de busca obtendo resultados de alta qualidade. Os algoritmos híbridos propostos são aplicados tanto no problema de job shop com tempo de processamento preciso, como nos problemas de job shop com tempo de processamento incerto. No caso de problema com tempo de processamento incerto, os algoritmos visam encontrar um conjunto diversificado de escalonamentos com alto grau de possibilidade de serem ótimos / Abstract: This work proposes new hybrid approaches based on techniques of bio-inspired computing for the Job Shop scheduling problem. As the job shop scheduling problem is NP-hard and there is no exact algorithm capable of solving all kinds of this problem. Usually it is necessary to elaborate more sophisticated methods of resolution to overcome this high complexity. Thus, in this work we propose hybrid approaches based on memetic algorithm and ant colony optimization algorithm in order to explore the search space in an efficient manner and obtain high quality results. The proposed hybrid algorithms are applied in both the job shop scheduling problem with precise processing time, as in job shop scheduling problems with uncertain processing time. In the case of problem with uncertain processing time, the algorithms obtain a diversified set of schedules with high possibility of being optimal / Doutorado / Automação / Doutor em Engenharia Elétrica

Dans une optique de maximisation de la production et de réduction des coûts d’installation, de maintenance et d’entretien des trackers solaires, qui permettent d’orienter les modules photovoltaïques à haute concentration (HCPV), ces travaux de thèse se focalisent sur l’amélioration de la précision et la réduction du coût de la stratégie de génération de la trajectoire du tracker. Dans un premier temps, un simulateur de tracker HCPV est développé offrant une étude de l’influence de la performance du suivi du soleil sur la production des modules HCPV, permettant ainsi une étude et une comparaison des stratégies de génération de trajectoires. Le simulateur est basé sur un modèle comportemental de module HCPV monté sur tracker permettant de prédire la puissance maximale du module HCPV en fonction de l’erreur de position du tracker face au soleil, de l’ensoleillement direct et de la température. Une première stratégie de commande dite de référence a été implémentée sur ce simulateur. C’est une commande hybride qui repose sur un viseur solaire pour corriger l’erreur de poursuite par un calcul astronomique. Ensuite, afin d’améliorer les performances et de réduire les coûts de cette stratégie, une nouvelle approche sans capteur est développée en se basant sur une méthode d’optimisation du gradient de puissance pour la génération de la trajectoire du tracker. Une étude complémentaire est également exposée afin de mettre en évidence des algorithmes de recherche de la puissance maximale (MPPT) pouvant offrir des temps de réponse suffisamment rapides pour ne pas affecter la qualité de l’évaluation du gradient de puissance. Dans ce contexte, une commande MPPT P&O améliorée par un réseau de neurones à complexité réduite est proposée, assurant un compromis entre précision, simplicité et rapidité / This work focuses on improving the accuracy and on reducing the cost of the tracker generating trajectory strategy, in order to maximize the production and to reduce the installation and the maintenance cost of a solar tracker orienting high concentrated photovoltaic modules (HCPV). Initially, we propose a behavioral modeling of the HCPV module mounted on a dual axis tracker in order to study the influence of the tracking performance on the module power production. Then, this simulator can be used to test control strategies and to compare their performance. Firstly, a classical control strategy is implemented in the simulator. It is based on a hybrid control operating an astronomical calculation to follow the sun path, and a sun sensor to correct the tracking error. A sensorless strategy is proposed in this work to reduce the cost of the HCPV tracker control. This strategy is based on a gradient optimization algorithm to generate the tracker trajectory and to catch the sun path. Tested on the simulator, this strategy presents the same accuracy as the classical strategy while being less costly. The last study proposed in this thesis work concerns maximum power point tracking (MPPT) algorithms, in order to respond to a given problem relating to the practical implementation of gradient algorithm. In this context, we propose an original optimization of the P&O MPPT control with a neural network algorithm leading to a significant reduction of the computational cost required to train it. This approach, which is ensuring a good compromise between accuracy and complexity is sufficiently fast to not affect the quality of the evaluation of the gradient.

The objective of the cellular manufacturing is to simplify the management of the<p>manufacturing industries. In regrouping the production of different parts into clusters,<p>the management of the manufacturing is reduced to manage different small<p>entities. One of the most important problems in the cellular manufacturing is the<p>design of these entities called cells. These cells represent a cluster of machines that<p>can be dedicated to the production of one or several parts. The ideal design of a<p>cellular manufacturing is to make these cells totally independent from one another,<p>i.e. that each part is dedicated to only one cell (i.e. if it can be achieved completely<p>inside this cell). The reality is a little more complex. Once the cells are created,<p>there exists still some traffic between them. This traffic corresponds to a transfer of<p>a part between two machines belonging to different cells. The final objective is to<p>reduce this traffic between the cells (called inter-cellular traffic).<p>Different methods exist to produce these cells and dedicated them to parts. To<p>create independent cells, the choice can be done between different ways to produce<p>each part. Two interdependent problems must be solved:<p>• the allocation of each operation on a machine: each part is defined by one or<p>several sequences of operations and each of them can be achieved by a set of<p>machines. A final sequence of machines must be chosen to produce each part.<p>• the grouping of each machine in cells producing traffic inside and outside the<p>cells.<p>In function of the solution to the first problem, different clusters will be created to<p>minimise the inter-cellular traffic.<p>In this thesis, an original method based on the grouping genetic algorithm (Gga)<p>is proposed to solve simultaneously these two interdependent problems. The efficiency<p>of the method is highlighted compared to the methods based on two integrated algorithms<p>or heuristics. Indeed, to form these cells of machines with the allocation<p>of operations on the machines, the used methods permitting to solve large scale<p>problems are generally composed by two nested algorithms. The main one calls the<p>secondary one to complete the first part of the solution. The application domain goes<p>beyond the manufacturing industry and can for example be applied to the design of<p>the electronic systems as explained in the future research.<p> / Doctorat en Sciences de l'ingénieur / info:eu-repo/semantics/nonPublished

I extended an artificial life platform called GReaNs (the name stands for Gene Regulatory evolving artificial Networks) to explore the evolutionary abilities of biologically inspired Spiking Neural Network (SNN) model. The encoding of SNNs in GReaNs was inspired by the encoding of gene regulatory networks. As proof-of-principle, I used GReaNs to evolve SNNs to obtain a network with an output neuron which generates a predefined spike train in response to a specific input. Temporal pattern recognition was one of the main tasks during my studies. It is widely believed that nervous systems of biological organisms use temporal patterns of inputs to encode information. The learning technique used for temporal pattern recognition is not clear yet. I studied the ability to evolve spiking networks with different numbers of interneurons in the absence and the presence of noise to recognize predefined temporal patterns of inputs. Results showed, that in the presence of noise, it was possible to evolve successful networks. However, the networks with only one interneuron were not robust to noise. The foraging behaviour of many small animals depends mainly on their olfactory system. I explored whether it was possible to evolve SNNs able to control an agent to find food particles on 2-dimensional maps. Using ring rate encoding to encode the sensory information in the olfactory input neurons, I managed to obtain SNNs able to control an agent that could detect the position of the food particles and move toward it. Furthermore, I did unsuccessful attempts to use GReaNs to evolve an SNN able to control an agent able to collect sound sources from one type out of several sound types. Each sound type is represented as a pattern of different frequencies. In order to use the computational power of neuromorphic hardware, I integrated GReaNs with the SpiNNaker hardware system. Only the simulation part was carried out using SpiNNaker, but the rest steps of the genetic algorithm were done with GReaNs.

Nonnegative matrix factorization (NMF) is an important data-analysis problem which concerns factoring a given d n matrix A with nonnegative entries into matrices B and C where B and C are d k and k n with nonnegative entries. It has numerous applications including Object recognition, Topic Modelling, Hyper-spectral imaging, Music transcription etc. In general, NMF is intractable and several heuristics exists to solve the problem of NMF. Recently there has been interest in investigating conditions under which NMF can be tractably recovered. We note that existing attempts make unrealistic assumptions and often the associated algorithms tend to be not scalable. In this thesis, we make three major contributions: First, we formulate a model of NMF with assumptions which are natural and is a substantial weakening of separability. Unlike requiring a bound on the error in each column of (A BC) as was done in much of previous work, our assumptions are about aggregate errors, namely spectral norm of (A BC) i.e. jjA BCjj2 should be low. This is a much weaker error assumption and the associated B; C would be much more resilient than existing models. Second, we describe a robust polynomial time SVD-based algorithm, UTSVD, with realistic provable error guarantees and can handle higher levels of noise than previous algorithms. Indeed, experimentally we show that existing NMF models, which are based on separability assumptions, degrade much faster than UTSVD, in the presence of noise. Furthermore, when the data has dominant features, UTSVD significantly outperforms existing models. On real life datasets we again see a similar outperformance of UTSVD on clustering tasks. Finally, under a weaker model, we prove a robust version of uniqueness of NMF, where again, the word \robust" refers to realistic error bounds.

The analysis and optimum design of stiffened, shear webs in aircraft structures is addressed. The post-buckling behaviour of the webs is assessed using the interactive algorithm developed by Grisham. This method requires only linear finite element analyses, while convergence is typically achieved in as few as five iterations. The Grisham algorithm is extensively compared with empirical analysis methods previously used for aircraft structures and also with a refined, non-linear quasi-static finite element analysis. The Grisham algorithm provides for both compressive buckling in two directions as well as shear buckling, and overcomes some of the conservatism inherent in conventional methods of analysis. In addition, the method is notably less expensive than a complete non-linear finite element analysis, even though global collapse cannot be predicted. While verification of the analysis methodology is the main focus of the stud, an initial investigation into optimization is also made. In optimizing stiffened thin walled structures, the Grisham algorithm is combined with a genetic algorithm. Allowable stress constraints are accommodated using a simple penalty formulation. / Dissertation (MEng (Mechanical and Aeronautical Engineering))--University of Pretoria, 2006. / Mechanical and Aeronautical Engineering / unrestricted