Composites have now revolutionized most industries, like aerospace, marine, electrical, transportation, and have proved to be a worthy alternative to other traditional materials. However for a further comprehensive usage, the tailorability of hybrid composites according to the specific application needs on a large-scale production basis is required. In this regard, one of the major fundamental research fields here involves a technology development based on the multiaxial warp-knitting technique for the production of bionic-inspired and application-specific textile preforms that are force compliant and exhibit multi-material design. This article presents a newly developed yarn (warp) path manipulation unit for multiaxial warp-knitting machines that enables a targeted production of customized textile preforms with the above characteristics. The technological development cycle and their experimental validation to demonstrate the feasibility of new technology through production of some patterns for different field of applications are then discussed.

Les travaux présentés dans ce manuscrit de thèse s’articulent autour de la synthèse de nouvelles architectures moléculaires tridimensionnelles et de l’évaluation de ces architectures en tant que canaux ioniques synthétiques capables de promouvoir le transport transmembranaire d’ions. La première partie concerne la mise au point d’une voie d’accès à ces édifices ayant comme plateforme centrale une brique naphtothiophène, aromatique ou partiellement hydrogénée, ainsi que l’étude prospective de la conversion de ces architectures en plateforme hélicoïdale. Les propriétés de ces édifices sont étudiées à l’état solide et par voie de calculs théoriques, permettant de mettre en avant des informations quant à la topologie globale adoptée ainsi que la compréhension de certaines réactivités observées. Une seconde partie est dédiée à la fonctionnalisation de ces édifices en molécules présentant des propriétés amphiphiles puis à l’étude de la capacité de ces dernières à s’insérer dans une bicouche lipidiques modèle afin de promouvoir le transport d’ions à travers la membrane via la formation de canaux ioniques dits synthétiques. En outre, des études alliant des analyses de spectrométrie de masse et des calculs théoriques sont présentés afin de comprendre les interactions intervenant dans le processus de transport d’ions à travers la membrane lipidique. / The work presented in this manuscript is dealing with the synthesis of new three-dimensional molecular architectures and their evaluation as synthetic ion channels capable of promoting ion transmembrane transport. The first part aims at developing a straightforward approach to the synthesis of novel architectures based on a naphthothiophene platform, aromatic or partially hydrogenated, as well as the development of a strategy the convert 9-arylnaphthothiophene architectures into helical platforms. The properties of these molecules were studied in the solid state and were completed by theoretical calculations to highlight global topologies adopted. Theoretical calculations allowed us to understanding some reactivities observed. A second part is dedicated firstly to the functionalisation of these molecular architectures into amphiphilic molecules and secondly to study their abilities to insert themselves into a model bilayer lipid membrane by forming channels. Besides, in order to gain a better understanding of the interactions in play in the process, mass spectrometry analysis combined to theoretical calculations were set up.

Cette thèse a été effectuée dans le cadre du projet européen Koroibot dont l'objectif est le développement d'algorithmes de marche avancés pour les robots humanoïdes. Dans le but de contrôler les robots d'une manière sûre et efficace chez les humains, il est nécessaire de comprendre les règles, les principes et les stratégies de l'homme lors de la locomotion et de les transférer à des robots. L'objectif de cette thèse est d'étudier et d'identifier les stratégies de locomotion humaine et créer des algorithmes qui pourraient être utilisés pour améliorer les capacités du robot. La contribution principale est l'analyse sur les principes de piétons qui guident les stratégies d'évitement des collisions. En particulier, nous observons comment les humains adapter une tâche de locomotion objectif direct quand ils ont à interférer avec un obstacle en mouvement traversant leur chemin. Nous montrons les différences entre la stratégie définie par les humains pour éviter un obstacle non-collaboratif et la stratégie pour éviter un autre être humain, et la façon dont les humains interagissent avec un objet si se déplaçant en manier simil à l'humaine. Deuxièmement, nous présentons un travail effectué en collaboration avec les neuroscientifiques de calcul. Nous proposons une nouvelle approche pour synthétiser réalistes complexes mouvements du robot humanoïde avec des primitives de mouvement. Trajectoires humaines walking-to-grasp ont été enregistrés. L'ensemble des mouvements du corps sont reciblées et proportionnée afin de correspondre à la cinématique de robots humanoïdes. Sur la base de cette base de données des mouvements, nous extrayons les primitives de mouvement. Nous montrons que ces signaux sources peuvent être exprimées sous forme de solutions stables d'un système dynamique autonome, qui peut être considéré comme un système de central pattern generators (CPGs). Sur la base de cette approche, les stratégies réactives walking-to-grasp ont été développés et expérimenté avec succès sur le robot humanoïde HRP-2 au LAAS-CNRS. Dans la troisième partie de la thèse, nous présentons une nouvelle approche du problème de pilotage d'un robot soumis à des contraintes non holonomes par une porte en utilisant l'asservissement visuel. La porte est représentée par deux points de repère situés sur ses supports verticaux. La plan géométric qui a été construit autour de la porte est constituée de faisceaux de hyperboles, des ellipses et des cercles orthogonaux. Nous montrons que cette géométrie peut être mesurée directement dans le plan d'image de la caméra et que la stratégie basée sur la vision présentée peut également être lié à l'homme. Simulation et expériences réalistes sont présentés pour montrer l'efficacité de nos solutions. / This thesis has been done within the framework of the European Project Koroibot which aims at developing advanced algorithms to improve the humanoid robots locomotion. It is organized in three parts. With the aim of steering robots in a safe and efficient manner among humans it is required to understand the rules, principles and strategies of human during locomotion and transfer them to robots. The goal of this thesis is to investigate and identify the human locomotion strategies and create algorithms that could be used to improve robot capabilities. A first contribution is the analysis on pedestrian principles which guide collision avoidance strategies. In particular, we observe how humans adapt a goal-direct locomotion task when they have to interfere with a moving obstacle crossing their way. We show differences both in the strategy set by humans to avoid a non-collaborative obstacle with respect to avoid another human, and the way humans interact with an object moving in human-like way. Secondly, we present a work done in collaboration with computational neuroscientists. We propose a new approach to synthetize realistic complex humanoid robot movements with motion primitives. Human walking-to-grasp trajectories have been recorded. The whole body movements are retargeted and scaled in order to match the humanoid robot kinematics. Based on this database of movements, we extract the motion primitives. We prove that these sources signals can be expressed as stable solutions of an autonomous dynamical system, which can be regarded as a system of coupled central pattern generators (CPGs). Based on this approach, reactive walking-to-grasp strategies have been developed and successfully experimented on the humanoid robot HRP at LAAS-CNRS. In the third part of the thesis, we present a new approach to the problem of vision-based steering of robot subject to non-holonomic constrained to pass through a door. The door is represented by two landmarks located on its vertical supports. The planar geometry that has been built around the door consists of bundles of hyperbolae, ellipses, and orthogonal circles. We prove that this geometry can be directly measured in the camera image plane and that the proposed vision-based control strategy can also be related to human. Realistic simulation and experiments are reported to show the effectiveness of our solutions.

Background: The world is facing environmental issues such as pollution, natural disasters, and climate change. A main cause of this is human activities such as urbanization and mining of materials. As the understanding of sustainability increases, new regulations from governments arise along with a surge in demand for sustainable solutions from customers. These factors will trigger a technological shift that companies need to tackle in order to stay competitive. Objectives: The aim of this research is to examine the barriers towards implementing bio-inspired methods for innovation. This will be done by studying innovators in relevant industries. Research question: What are the challenges with a large scale implementation of bio-inspired methods for innovation? Method: This study will be carried out by interviewing people working within the transportation and heavy construction industries. The candidates will, on a daily basis, work with innovation and thus contribute with their knowledge and reflections. The study will use semi-structured interviews followed by a thematic analysis to get results and draw conclusions. Results: The study shows that using methods for innovation is a common act. Furthermore, the candidates were asked how they integrate sustainability into their development. In general, the candidates had no unified answer to this question. However, several examples, such as lowering emissions, not buying from corrupt suppliers, and maintaining profitability, were mentioned. Moreover, the candidates states that the company works with identifying new methods for innovation. When asked about bio-inspired methods, most candidates were positive about an implementation. From another perspective, some candidates who work with biomimicry state the opposite. Conclusions: Using bio-inspired methods for innovation might be of interest, however, such an implementation takes long time due to organizational resistance. To understand this further, more industries and perspectives has to be studied.

[pt] Redes neurais profundas são responsáveis pelo grande progresso em diversas tarefas perceptuais, especialmente nos campos da visão computacional,reconhecimento de fala e processamento de linguagem natural. Estes resultados produziram uma mudança de paradigma nas técnicas de reconhecimentode padrões, deslocando a demanda do design de extratores de característicaspara o design de arquiteturas de redes neurais. No entanto, o design de novas arquiteturas de redes neurais profundas é bastante demandanteem termos de tempo e depende fortemente da intuição e conhecimento de especialistas,além de se basear em um processo de tentativa e erro. Neste contexto, a idea de automatizar o design de arquiteturas de redes neurais profundas tem ganhado popularidade, estabelecendo o campo da busca por arquiteturas neurais(NAS - Neural Architecture Search). Para resolver o problema de NAS, autores propuseram diversas abordagens envolvendo o espaço de buscas, a estratégia de buscas e técnicas para mitigar o consumo de recursos destes algoritmos. O Q-NAS (Quantum-inspired Neural Architecture Search) é uma abordagem proposta para endereçar o problema de NAS utilizando um algoritmo evolucionário com inspiração quântica como estratégia de buscas. Este método foi aplicado de forma bem sucedida em classificação de imagens, superando resultados de arquiteturas de design manual nos conjuntos de dados CIFAR-10 e CIFAR-100 além de uma aplicação de mundo real na área da sísmica. Motivados por este sucesso, propõe-se nesta Dissertação o SegQNAS (Quantum-inspired Neural Architecture Search applied to Semantic Segmentation), uma adaptação do Q-NAS para a tarefa de segmentação semântica. Diversos experimentos foram realizados com objetivo de verificar a aplicabilidade do SegQNAS em dois conjuntos de dados do desafio Medical Segmentation Decathlon. O SegQNAS foi capaz de alcançar um coeficiente de similaridade dice de 0.9583 no conjunto de dados de baço, superando os resultados de arquiteturas tradicionais como U-Net e ResU-Net e atingindo resultados comparáveis a outros trabalhos que aplicaram NAS a este conjunto de dados, mas encontrando arquiteturas com muito menos parãmetros. No conjunto de dados de próstata, o SegQNAS alcançou um coeficiente de similaridade dice de 0.6887 superando a U-Net, ResU-Net e o trabalho na área de NAS que utilizamos como comparação. / [en] Deep neural networks are responsible for great progress in performance for several perceptual tasks, especially in the fields of computer vision, speech recognition, and natural language processing. These results produced a paradigm shift in pattern recognition techniques, shifting the demand from feature extractor design to neural architecture design. However, designing novel deep neural network architectures is very time-consuming and heavily relies on experts intuition, knowledge, and a trial and error process. In that context, the idea of automating the architecture design of deep neural networks has gained popularity, establishing the field of neural architecture search (NAS). To tackle the problem of NAS, authors have proposed several approaches regarding the search space definition, algorithms for the search strategy, and techniques to mitigate the resource consumption of those algorithms. Q-NAS (Quantum-inspired Neural Architecture Search) is one proposed approach to address the NAS problem using a quantum-inspired evolutionary algorithm as the search strategy. That method has been successfully applied to image classification, outperforming handcrafted models on the CIFAR-10 and CIFAR-100 datasets and also on a real-world seismic application. Motivated by this success, we propose SegQNAS (Quantum-inspired Neural Architecture Search applied to Semantic Segmentation), which is an adaptation of Q-NAS applied to semantic segmentation. We carried out several experiments to verify the applicability of SegQNAS on two datasets from the Medical Segmentation Decathlon challenge. SegQNAS was able to achieve a 0.9583 dice similarity coefficient on the spleen dataset, outperforming traditional architectures like U-Net and ResU-Net and comparable results with a similar NAS work from the literature but with fewer parameters network. On the prostate dataset, SegQNAS achieved a 0.6887 dice similarity coefficient, also outperforming U-Net, ResU-Net, and outperforming a similar NAS work from the literature.

No description available.

The necessity for clean and sustainable energy has shifted the energy sector’s interest in renewable energy sources. Photovoltaics (PV) is the most popular renewable energy source because the sun is ubiquitous. However, several discrepancies exist in a PV system when implemented for real-world applications. Among several other existing problems related to Photovoltaics, in this work, we deal with maximum power point tracking (MPPT) under Partial Shading (PS) conditions. MPPT is a mechanism formulated as an optimization problem adjusting the PV to deliver the maximum power to the load. Under full insolation conditions, varying solar panel temperatures, and different loads MPPT problem is a convex optimization problem. However, when the PV’s surface is partially shaded, multiple power peaks are created in the power versus voltage (P-V) curve making MPPT non-convex.

Furthering our understanding of the physics of flapping flight has the potential to benefit the field of micro air vehicles. Advancements in micro air vehicles can benefit applications such as surveillance, reconnaissance, and search and rescue. In this research, flapping kinematics of a ladybug was explored using a direct linear transformation. A flapping mechanism design is presented that was capable of executing ladybug or other species-specific kinematics. The mechanism was based on a differential gear design, had two wings, and could flap in harsh environments. This mechanism served as a test bed for force analysis and optimization studies. The first study was based on a Box-Behnken screening design to explore wing kinematic parameter design space and manually search in the direction of flapping kinematics that optimized the objective of maximum combined lift and thrust. The second study used a Box-Behnken screening design to build a response surface. Using gradient-based techniques, this surface was optimized for maximum combined lift and thrust. Box-Behnken design coupled with response surface methodology was an efficient method for exploring the mechanism force response. Both methods for optimization were capable of successfully improving lift and thrust force outputs. The incorporation of the results of these studies will aid in the design of more efficient micro air vehicles and with the ultimate goal of leading to a better understanding of flapping wing aerodynamics and the development of aerodynamic models.

[pt] As redes neurais deep são modelos poderosos e flexíveis, que ganharam destaque na comunidade científica na última década. Para muitas tarefas, elas até superam o desempenho humano. Em geral, para obter tais resultados, um especialista despende tempo significativo para projetar a arquitetura neural, com longas sessões de tentativa e erro. Com isso, há um interesse crescente em automatizar esse processo. Novos métodos baseados em técnicas como aprendizado por reforço e algoritmos evolutivos foram apresentados como abordagens para o problema da busca de arquitetura neural (NAS - Neural Architecture Search), mas muitos ainda são algoritmos de alto custo computacional. Para reduzir esse custo, pesquisadores sugeriram limitar o espaço de busca, com base em conhecimento prévio. Os algoritmos evolutivos de inspiração quântica (AEIQ) apresentam resultados promissores em relação à convergência mais rápida. A partir dessa idéia, propõe-se o Q-NAS: um AEIQ para buscar redes deep através da montagem de subestruturas. O Q-NAS também pode evoluir alguns hiperparâmetros numéricos, o que é um primeiro passo para a automação completa. Experimentos com o conjunto de dados CIFAR-10 foram realizados a fim de analisar detalhes do Q-NAS. Para muitas configurações de parâmetros, foram obtidos resultados satisfatórios. As melhores acurácias no CIFAR-10 foram de 93,85 porcento para uma rede residual e 93,70 porcento para uma rede convolucional, superando modelos elaborados por especialistas e alguns métodos de NAS. Incluindo um esquema simples de parada antecipada, os tempos de evolução nesses casos foram de 67 dias de GPU e 48 dias de GPU, respectivamente. O Q-NAS foi aplicado ao CIFAR-100, sem qualquer ajuste de parâmetro, e obteve 74,23 porcento de acurácia, similar a uma ResNet com 164 camadas. Por fim, apresenta-se um estudo de caso com dados reais, no qual utiliza-se o Q-NAS para resolver a tarefa de classificação sísmica. Em menos de 8,5 dias de GPU, o Q-NAS gerou redes com 12 vezes menos pesos e maior acurácia do que um modelo criado especialmente para esta tarefa. / [en] Deep neural networks are powerful and flexible models that have gained the attention of the machine learning community over the last decade. For a variety of tasks, they can even surpass human-level performance. Usually, to reach these excellent results, an expert spends significant time designing the neural architecture, with long trial and error sessions. In this scenario, there is a growing interest in automating this design process. To address the neural architecture search (NAS) problem, authors have presented new methods based on techniques such as reinforcement learning and evolutionary algorithms, but the high computational cost is still an issue for many of them. To reduce this cost, researchers have proposed to restrict the search space, with the help of expert knowledge. Quantum-inspired evolutionary algorithms present promising results regarding faster convergence. Motivated by this idea, we propose Q-NAS: a quantum-inspired algorithm to search for deep networks by assembling substructures. Q-NAS can also evolve some numerical hyperparameters, which is a first step in the direction of complete automation. We ran several experiments with the CIFAR-10 dataset to analyze the details of the algorithm. For many parameter settings, Q-NAS was able to achieve satisfactory results. Our best accuracies on the CIFAR-10 task were 93.85 percent for a residual network and 93.70 percent for a convolutional network, overcoming hand-designed models, and some NAS works. Considering the addition of a simple early-stopping mechanism, the evolution times for these runs were 67 GPU days and 48 GPU days, respectively. Also, we applied Q-NAS to CIFAR-100 without any parameter adjustment, reaching an accuracy of 74.23 percent, which is comparable to a ResNet with 164 layers. Finally, we present a case study with real datasets, where we used Q-NAS to solve the seismic classification task. In less than 8.5 GPU days, Q-NAS generated networks with 12 times fewer weights and higher accuracy than a model specially created for this task.

No description available.