This thesis presents an automatic animation system for jellyfish that is based on a physical simulation of the organism and its surrounding fluid. Our goal is to explore the unusual style of locomotion, namely jet propulsion, which is utilized by jellyfish. The organism achieves this propulsion by contracting its body, expelling water, and propelling itself forward. The organism then expands again to refill itself with water for a subsequent stroke. We endeavor to model the thrust achieved by the jellyfish, and also the evolution of the organism's geometric configuration. <p> We restrict our discussion of locomotion to fully grown adult jellyfish, and we restrict our study of locomotion to the resonant gait, which is the organism's most active mode of locomotion, and is characterized by a regular contraction rate that is near one of the creature's resonant frequencies. We also consider only species that are axially symmetric, and thus are able to reduce the dimensionality of our model. We can approximate the full 3D geometry of a jellyfish by simulating a 2D slice of the organism. This model reduction yields plausible results at a lower computational cost. From the 2D simulation, we extrapolate to a full 3D model. To prevent our extrapolated model from being artificially smooth, we give the final shape more variation by adding noise to the 3D geometry. This noise is inspired by empirical data of real jellyfish, and also by work with continuous noise functions from the graphics community. <p> Our 2D simulations are done numerically with ideas from the field of computational fluid dynamics. Specifically, we simulate the elastic volume of the jellyfish with a spring-mass system, and we simulate the surrounding fluid using the semi-Lagrangian method. To couple the particle-based elastic representation with the grid-based fluid representation, we use the immersed boundary method. We find this combination of methods to be a very efficient means of simulating the 2D slice with a minimal compromise in physical accuracy.

[pt] Esse trabalho tem como objetivo desenvolver um modelo determinístico transiente de previsão de escorregamentos planares em encostas, para escalas em nível de bacia hidrográfica (1:2000 a 1:5000). No modelo são aplicadas as teorias de Green-Ampt (1934) e de O Loughlin (1986), essa última utilizada no programa SHALSTAB (MONTGOMERY e DIETRICH, 1994), além da teoria talude infinito 2D e 3D. Também são considerados nas análises a não saturação do solo e os efeitos da vegetação. O evento estudado para aplicação e teste do modelo refere-se ao ocorrido em 1996, nas bacias do Quitite e Papagaio em Jacarepaguá, Zona Oeste do Rio de Janeiro. Além do mapeamento do fator de segurança nas bacias, também são gerados mapas com o escoamento superficial acumulado, e tenta-se correlacionar ambos com as cicatrizes que ocorreram no caso de estudo. / [en] The aim of this work is to develop a physically-based transient model for the prediction of planar landslides in natural slopes. The application scale of the model is for a hydrographic basin (1:2000 to 1:5000). The theories of Green-Ampt (1934) and O Loughlin (1986), the second one used in the SHALSTAB program (MONTGOMERY e DIETRICH, 1994), and the infinite slope 2D and 3D are used in the model. The effect of the unsaturation and the vegetation is also considered in the analysis. The case study for the test of the model is the 1996 event that took place in the Quitite and Papagaio basins, in Jacarepaguá, Zona Oeste of Rio de Janeiro. Besides mapping the safety factor in the basins, maps of the accumulated runoff were also generated. This work also tries to correlate the runoff as another factor that caused the landslides.

This thesis describes a system for dealing with free surface fluid simulations, and the components needed in order to construct such a system. It builds upon recent research, but in a computer graphics context the amount of available literature is limited and difficult to implement. Because of this, the text aims at providing a solid foundation of the mathematics needed, at explaining in greater detail the steps needed to solve the problem, and lastly at improving some aspects of the animation process as it has been described in earlier works. The aim of the system itself is to provide visually plausible renditions of animated fluids in three dimensions in a manner that allows it to be usable in a visual effects production context. The novel features described include a generalized interaction layer providing greater control to artists, a new way of dealing with moving objects that interact with the fluid and a method for adding source and drain capabilities.

La Théorie de la Relativité (TRR), est une théorie particulièrement contre-intuitive dont les implications sont inaccessibles à l'expérience sensible humaine ; ce qui pose un certain nombre de difficultés de compréhension aux étudiants. Cependant, la Réalité Virtuelle (RV) offre une approche intéressante en permettant à un utilisateur d'être immergé et d'interagir dans un monde virtuel où la vitesse de la lumière est ramenée à 1 m/s. Les phénomènes relativistes deviennent ainsi directement accessibles par ses sens. Cette caractéristique, point départ de nos travaux, permet alors d’appréhender les phénomènes relativistes par une expérience « par la pratique ». L'enjeu de notre travail porte plus précisément sur la définition de moyens et de méthodes intégrés dans une plate-forme immersive permettant d'appréhender les phénomènes relativistes. Dans ce contexte, nous proposons, tout d’abord, des méthodes novatrices pour simuler les phénomènes relativistes sur un nombre quelconque d'objets en mouvement arbitraire et tenant compte de la dynamique relativiste des objets dans la scène, notamment durant leurs interactions. Nous nous focalisons sur les effets qui déforment les objets vus par l'observateur, à savoir le délai de propagation des photons, la relativité des longueurs et l'effet d'aberration. Nous définissons ensuite des méthodes pour intégrer une simulation relativiste dans un environnement immersif basé intrinsèquement sur un monde newtonien. Nous proposons également une plate-forme expérimentale dans laquelle sont intégrées des méthodes d'interaction utilisées pour mettre en scène un « jeu sérieux », ici un billard relativiste. Enfin, nous démontrons la portée de notre outil expérimental par deux voies : l'une concerne l'utilisation de l'application dans des évaluations de didactique et l'autre concerne un exemple d'extension de l'outil pour mettre en lumière un autre aspect de la Physique relativiste : la relation entre vitesse et énergie. / The Theory of Special Relativity (TSR) is a particularly counterintuitive theory. Its implications are, by nature, out of reach by human experience. Therefore we cannot perceive its effects directly, thus raising problems of comprehension for the students confronted to it. However, Virtual Reality (VR) enables us to overcome this limitation by immersing a user into a world where the velocity of light is reduced to 1 m/s. As a result, the relativistic phenomena become directly perceivable through our senses. This possibility, which is the cornerstone of our work, brings a unique way to apprehend the relativistic phenomenon trough a "hands-on" experiment.In this context, we propose, first, innovative methods to include relativistic effects in simulation containing any number of objects moving in an arbitrary direction and velocity and taking into account the relativistic dynamics of the objects, including object-to-object interaction. We focused on the relativistic phenomenon involved in the deformation of objects: the delay of propagation of the photons from the light source to the observer, as well as the relativity of length and the aberration of light. We describe, second, methods to integrate the simulation techniques, previously introduced, into an immersive environment intrinsically based on Newtonian physics. We also provide interaction methods and a concrete application in a serious game framework: a relativistic carom billiard. Finally, we demonstrate the possibilities of our platform are demonstrated in two ways: one tackles usage in the context of learning evaluation and the other is an extension of the tool to access new pieces of information relevant to TSR, such as the force profile used to launch an object with a relativistic velocity.

Modern machine learning methods, utilising neural networks, require a lot of training data. Data gathering and preparation has thus become a major bottleneck in the machine learning pipeline and researchers often use large public datasets to conduct their research (such as the ImageNet [1] or MNIST [2] datasets). As these methods begin being used in industry, these challenges become apparent. In factories objects being produced are often unique and may even involve trade secrets and patents that need to be protected. Additionally, manufacturing may not have started yet, making real data collection impossible. In both cases a public dataset is unlikely to be applicable. One possible solution, investigated in this thesis, is synthetic data generation. Synthetic data generation using physically based rendering was tested for unsupervised anomaly detection on a 3D printed block. A small image dataset was gathered of the block as control and a data generation model was created using its CAD model, a resource most often available in industrial settings. The data generation model used randomisation to reduce the domain shift between the real and synthetic data. For testing the data, autoencoder models were trained, both on the real and synthetic data separately and in combination. The material of the block, a white painted surface, proved challenging to reconstruct and no significant difference between the synthetic and real data could be observed. The model trained on real data outperformed the models trained on synthetic and the combined data. However, the synthetic data combined with the real data showed promise with reducing some of the bias intentionally introduced in the real dataset. Future research could focus on creating synthetic data for a problem where a good anomaly detection model already exists, with the goal of transferring some of the synthetic data generation model (such as the materials) to a new problem. This would be of interest in industries where they produce many different but similar objects and could reduce the time needed when starting a new machine learning project.

In computer graphics applications, the choice and implementation of a rendering technique is crucial when targeting real-time performance. Traditionally, rasterization-based approaches have dominated the real-time sector. Other algorithms were simply too slow to compete on consumer graphics hardware. With the addition of hardware support for ray-intersection calculations on modern GPUs, hybrid ray tracing/rasterization and purely ray tracing approaches have become possible in real-time as well. Industry real-time graphics applications, namely games, have been exploring these different rendering techniques with great levels of success. The addition of ray tracing into the graphics developer’s toolkit has without a doubt increased what level of graphical fidelity is achievable in real-time. In this thesis, three rendering techniques are implemented in a custom rendering engine built on the Vulkan® Explicit API. Each technique represents a different family of modern real-time rendering algorithms. A largely rasterization-based method, a hybrid ray tracing/rasterization method, and a method solely using ray tracing. Both the hybrid and ray tracing exclusive approach rely on the ReSTIR algorithm for lighting calculations. Analysis of the performance and render quality of these approaches reveals the trade-offs incurred by each approach, alongside the performance viability of each in a real-time setting.

L'un des objectifs lors du développement d'un produit industriel est d'obtenir un prototype numérique valide et réaliste. Cette thèse a pour objectif d'améliorer la qualité des simulations dans le contexte d'un processus de production. Ces processus impliquent souvent un rendu de type "walkthrough", avec une géométrie fixe mais un déplacement continu de l'observateur. Nous nous intéresserons donc plus précisément aux méthodes de rendu physiquement réaliste de scènes complexes pour un scénario "walkthrough". Durant le rendu, l'utilisateur doit pouvoir mesurer précisément la radiance d'un point ou d'une zone donnée, ainsi que modifier en temps réel la puissance des sources lumineuses. Fondée sur la méthode du photon mapping, nos travaux montrent les modifications à apporter aux algorithmes afin d'améliorer à la fois la qualité des images et le temps de calcul du processus de rendu. / One of the goals when developing the product is to immediately obtain a real and valid prototype. This thesis provide new rendering methods to increase the quality of the simulations during the upstream work of the production pipeline. The latter usually requires a walkthrough rendering. Thus, we focuses on the physically-based rendering methods of complex scenes in walkthrough. During the rendering, the end-users must be able to measure the illuminate rates and to interactively modify the power of the light source to test different lighting ambiances. Based on the original photon mapping method, our work shows how some modifications can decrease the calculation time and improve the quality of the resulting images according to this specific context.

The rendering of participating media is an interesting and important problem without a simple solution. Yet even among the wide variety of participating media the clouds stand out as an especially difficult case, because of their properties that make their simulation even harder. The work presented in this thesis attempts to provide a solution to this problem, and moreover, to make the proposed method to work in interactive rendering speeds. The main design criteria in designing this method were its physical plausibility and maximal utilization of specific cloud properties which would help to balance the complex nature of clouds. As a result the proposed method builds on the well known photon mapping algorithm, but modifies it in several ways to obtain interactive and temporarily coherent results. This is further helped by designing the method in such a way which allows its implementation on contemporary GPUs, taking advantage of their massively parallel sheer computational power. We implement a prototype of the method in an application that renders a single realistic cloud in interactive framerates, and discuss possible extensions of the proposed technique that would allow its use in various practical industrial applications.

Cette thèse traite de la diffusion par des surfaces rugueuses monodimensionnelles. Les surfaces présentant des petites échelles de variations nécessitent une discrétisation fine pour représenter les effets de diffusion sur le champ diffracté, ce qui augmente les coûts numériques. Deux aspects sont considérés : la réduction de la taille du problème en construisant une condition aux limiteséquivalente traduisant les effets des variations rapides et la réduction du nombre d’itérations nécessaires pour résoudre le système linéaire issu de la méthode des moments par une méthode basée sur les sous-espaces de Krylov. En ce qui concerne la réduction de la taille du problème, une technique d’homogénéisation est utilisée pour transformer la condition aux limites posée sur lasurface rugueuse par des paramètres effectifs. Ces paramètres sont déterminés par des problèmes auxiliaires qui tiennent compte des échelles fines de la surface. Dans le cas de surfaces parfaitement métalliques, la procédure est appliquée en polarisation Transverse Magnétique (TM) et Transverse Électrique (TE). Une impédance équivalente de Léontovich d’ordre 1 est déduite.Le procédure est automatique et les ordres supérieurs sont dérivés pour la polarisation TM. La procédure d’homogénéisation est aussi appliquée pour des interfaces rugueuses séparant deux milieux diélectriques. En ce qui concerne la réduction du nombre d’itérations, un préconditionneur, basé sur des considérations physiques, est construit à partir des modes de Floquet. Bien que le préconditionneur soit initialement élaboré pour des surfaces périodiques, nous montrons qu’il est aussi efficace pour des surfaces tronquées éclairées par une onde plane. L’efficacité des deux aspects présentés dans cette thèse est numériquement illustrée pour des configurations d’intérêt. / This work is about the scattering by monodimensional rough surfaces. Surfaces presenting small scales of variations need a very refined mesh to finally capture the scattering field behaviour what increases the computational cost. Two aspects are considered : the reduction of the problemsize through an effective boundary condition incorporating the effect of rapid variations and the reduction of the number of iterations to solve the linear system arising from method of moments by a method based on Krylov subspace. Firstly, an homogenization process is used to convert the boundary condition on the rough interface into effective parameters. These parameters are determined by the solutions of auxiliary problems which involve the detailed profile of the interface. In the case of perfectly metallic surfaces, the process is applied to the E- and H-polarization and an Leontovich impedance of order 1 is deduced. The process is automatic and higher orders are derived for E-polarization. The homogenization process is also applied to dielectric rough interfaces. Secondly, a physically-based preconditioner is built with Floquet’s modes. Although the preconditioner has been designed for periodical surfaces, it was shown to be efficient in the case of truncated surfaces illuminated by a plane wave. The efficiency of both aspects is numerically illustrated for some configurations of interest.

Le modèle physique masses interactions est puissant pour la simulation de comportements dynamiques très divers et pour la production de mouvements expressifs, riches et d'une grande complexité. En revanche, une difficulté inhérente à ce type de formalisme pour la production d'images animées réside dans le fait que les masses ponctuelles n'ont pas de spatialité ; il est donc difficile de produire des séquences d'images animées par le rendu direct des masses ponctuelles décrivant le mouvement. D'une manière générale, il est donc nécessaire de développer des méthodes qui étendent la spatialité de ces masses ponctuelles pour compléter la chaîne de production d'images animées par modèle physique particulaire. Une méthode, proposée par le laboratoire ICA, répond à ce type de problématique en permettant d'étendre la spatialité des masses ponctuelles en considérant l'interaction physique entre ces masses et un milieu. Il s'agit d'une métaphore du procédé physique de la gravure. Celle ci a permis de produire des images animées convaincantes de divers phénomènes visuels. Nous présentons dans ce document un élargissement de cette méthode notamment au cas 3D, ainsi qu'à de nouveaux comportements. De plus, l'algorithme de cette méthode a été parallélisé, ce qui nous a permis d'obtenir des simulations calculées en temps réel en utilisant la puissance actuelle des cartes graphiques. Afin de maitriser au mieux les possibilités de la méthode, nous avons développé un logiciel comprenant une interface graphique manipulable et interactive permettant de modéliser avec aisance différents comportements. Cette méthode a été intégrée dans des installations interactives artistiques multi-sensorielles fournissant un comportement dynamique riche et configurable, tout en permettant une interaction en temps réel avec le spectateur. / Mass – Interaction physical modeling is a powerful formalism for the simulation of various dynamic behaviors and for the production of expressive, rich and complex motions. However, there is an inherent matter of this type of formalism for animation production, which resides on the fact that masses have no spatiality. Thus, it is difficult to produce animation sequences directly from rendering mass point describing the movement. It is then necessary to develop methods that extend the masses spatiality in order to complete the animation process. ICA Laboratory addressed the problem with a method based on the physical simulation of interaction between these masses and a dynamic milieu, according to the metaphor of engraving. We present in this document an extension of this method notably towards 3D and other effects. Besides, the parallel implementation on Graphic Cards (GPU) allowed obtaining real time simulation. An interactive graphical interface was also developed to facilitate the creation of different models. We used this process in multi-sensory interactive art installations for its rich and dynamic ability to create shape from motion and interact in real time with spectators.