No description available.

Demonstrating strength and agility on virtual and real humanoids has been an important goal in computer graphics and robotics. However, developing physics- based controllers for various agile motor skills requires a tremendous amount of prior knowledge and manual labor due to complex mechanisms of the motor skills. The focus of the dissertation is to develop a set of computational tools to expedite the design process of physics-based controllers that can execute a variety of agile motor skills on virtual and real humanoids. Instead of designing directly controllers real humanoids, this dissertation takes an approach that develops appropriate theories and models in virtual simulation and systematically transfers the solutions to hardware systems. The algorithms and frameworks in this dissertation span various topics from spe- cific physics-based controllers to general learning frameworks. We first present an online algorithm for controlling falling and landing motions of virtual characters. The proposed algorithm is effective and efficient enough to generate falling motions for a wide range of arbitrary initial conditions in real-time. Next, we present a robust falling strategy for real humanoids that can manage a wide range of perturbations by planning the optimal contact sequences. We then introduce an iterative learning framework to easily design various agile motions, which is inspired by human learn- ing techniques. The proposed framework is followed by novel algorithms to efficiently optimize control parameters for the target tasks, especially when they have many constraints or parameterized goals. Finally, we introduce an iterative approach for exporting simulation-optimized control policies to hardware of robots to reduce the number of hardware experiments, that accompany expensive costs and labors.

This thesis introduces locomotion synthesis methods for humanoid characters. Motion synthesis is an under-constrained problem that requires additional constraints beyond user inputs. Two main approaches to introducing additional constraints are physics-based and data-driven. Despite significant progress in the past 20 years, major difficulties still exist for both approaches. In general, building animation systems that are flexible to user requirements while keeping the synthesized motions plausible remain a challenging task. The methods introduced in this thesis, presented in two-parts, aim to allow animation systems to be more flexible to user demands without radically violating constraints that are important for maintaining plausibility. In the first part of the thesis, we address an important subproblem in physics-based animation --- controller synthesis for humanoid characters. We describe a method for optimizing the parameters of a physics-based controller for full-body, 3D walking. The objective function includes terms for power minimization, angular momentum minimization, and minimal head motion, among others. Together these terms produce a number of important features of natural walking, including active toe-off, near-passive knee swing, and leg extension during swing. We then extend the algorithm to optimize for robustness to uncertainty. Many unknown factors, such as external forces, control torques, and user control inputs, cannot be known in advance and must be treated as uncertain. Controller optimization entails optimizing the expected value of the objective function, which is computed by Monte Carlo methods. We demonstrate examples with a variety of sources of uncertainty and task constraints. The second part of this thesis deals with the data-driven approach and the problem of motion modeling. Defining suitable models for human motion data is non-trivial. Simple linear models are not expressive enough, while more complex models require setting many parameters and are difficult to learn with limited data. Using Bayesian methods, we demonstrate how the Gaussian process prior can be used to derive a kernelized version of multilinear models. The result is a locomotion model that takes advantage of training data addressed by multiple indices to improve generalization to unseen motions.

Animations in the game industry are considered a bottleneck today, and a way to speed up the development in this area is by letting the computer procedurally generate the animations. This research has explored blending physics with keyframed animations in games, to create interactive and responsive characters, and to find out how these animations are perceived by players. This way of animating characters in real time has lately been used in a blossoming game genre called fumblecore. By letting physics do part of animation in real time, unpredictable and unique animations can be created. Two demos of a fumblecore game was created and tested on players with various gaming backgrounds. One had a traditional keyframed animation, and the other had a physically-blended animation. The results showed that the majority of participants of the study preferred the physically-blended animation over the traditionally keyframed one.

Homogeneous Charge Compression Ignition (HCCI) Engines hold promises of being the next generation of internal combustion engines due to their ability to produce high thermal efficiencies and low emission levels. HCCI combustion is achieved through the auto-ignition of a compressed homogenous fuel-air mixture, thus making it a "fusion" between spark-ignition and compression-ignition engines. The main challenge in developing HCCI engines is the absence of a combustion trigger hence making it difficult to control its combustion timing. The aim of this research project is to model and control a natural gas HCCI engine. Since HCCI depends primarily on temperature and chemical composition of the mixture, Exhaust Gas Recirculation (EGR) is used to control ignition timing. In this research, a thermodynamical, physics-based nonlinear model is developed to capture the main features of the HCCI engine. In addition, the Modified Knock Integral Model (MKIM), used to predict ignition timing, is optimized. To validate the nonlinear model, ignition timing under varying conditions using the MKIM approach is shown to be in accordance with data acquired from a model developed using a sophisticated engine simulation program, GT-Power. Most control strategies are based on a linear model, therefore, the nonlinear model is linearized using the perturbation method. The linear model is validated by comparing its performance with the nonlinear model about a suitable operating point. The control of ignition timing can be defined as a regulation process where the goal is to force the nonlinear model to track a desired ignition timing by controlling the EGR ratio. Parameters from the linear model are used to determine the gains of the LQR controller. The performance of the controller is validated by implementing it on the nonlinear model and observing its ability to track the desired timing with 0.5% error within a certain operating range. To increase the operating range of the controller and reduce steady-state error, an integrator is added to the LQR. Finally, it is shown that the LQR controller is able to successfully reject disturbance, parameter variation, as well as noise.

This thesis introduces locomotion synthesis methods for humanoid characters. Motion synthesis is an under-constrained problem that requires additional constraints beyond user inputs. Two main approaches to introducing additional constraints are physics-based and data-driven. Despite significant progress in the past 20 years, major difficulties still exist for both approaches. In general, building animation systems that are flexible to user requirements while keeping the synthesized motions plausible remain a challenging task. The methods introduced in this thesis, presented in two-parts, aim to allow animation systems to be more flexible to user demands without radically violating constraints that are important for maintaining plausibility. In the first part of the thesis, we address an important subproblem in physics-based animation --- controller synthesis for humanoid characters. We describe a method for optimizing the parameters of a physics-based controller for full-body, 3D walking. The objective function includes terms for power minimization, angular momentum minimization, and minimal head motion, among others. Together these terms produce a number of important features of natural walking, including active toe-off, near-passive knee swing, and leg extension during swing. We then extend the algorithm to optimize for robustness to uncertainty. Many unknown factors, such as external forces, control torques, and user control inputs, cannot be known in advance and must be treated as uncertain. Controller optimization entails optimizing the expected value of the objective function, which is computed by Monte Carlo methods. We demonstrate examples with a variety of sources of uncertainty and task constraints. The second part of this thesis deals with the data-driven approach and the problem of motion modeling. Defining suitable models for human motion data is non-trivial. Simple linear models are not expressive enough, while more complex models require setting many parameters and are difficult to learn with limited data. Using Bayesian methods, we demonstrate how the Gaussian process prior can be used to derive a kernelized version of multilinear models. The result is a locomotion model that takes advantage of training data addressed by multiple indices to improve generalization to unseen motions.

The low-frequency electromagnetic compatibility (EMC) is an increasingly important aspect in the design of practical systems to ensure the functional safety and reliability of complex products. The opportunities for using numerical techniques to predict and analyze system’s EMC are therefore of considerable interest in many industries. As the first phase of study, a proper model, including all the details of the component, was required. Therefore, the advances in EMC modeling were studied with classifying analytical and numerical models. The selected model was finite element (FE) modeling, coupled with the distributed network method, to generate the model of the converter’s components and obtain the frequency behavioral model of the converter. The method has the ability to reveal the behavior of parasitic elements and higher resonances, which have critical impacts in studying EMI problems. For the EMC and signature studies of the machine drives, the equivalent source modeling was studied. Considering the details of the multi-machine environment, including actual models, some innovation in equivalent source modeling was performed to decrease the simulation time dramatically. Several models were designed in this study and the voltage current cube model and wire model have the best result. The GA-based PSO method is used as the optimization process. Superposition and suppression of the fields in coupling the components were also studied and verified. The simulation time of the equivalent model is 80-100 times lower than the detailed model. All tests were verified experimentally. As the application of EMC and signature study, the fault diagnosis and condition monitoring of an induction motor drive was developed using radiated fields. In addition to experimental tests, the 3DFE analysis was coupled with circuit-based software to implement the incipient fault cases. The identification was implemented using ANN for seventy various faulty cases. The simulation results were verified experimentally. Finally, the identification of the types of power components were implemented. The results show that it is possible to identify the type of components, as well as the faulty components, by comparing the amplitudes of their stray field harmonics. The identification using the stray fields is nondestructive and can be used for the setups that cannot go offline and be dismantled

Stiff, ultra-lightweight thermal-formed polyimide panels considered in this dissertation are examples of next generation gossamer structures that resolve some of the technology barriers of previous, membrane-dominated gossamer designs while maintaining their low mass and low stowage volume characteristics. The research involved statically and dynamically characterizing and modeling several of these panels to develop validated computer models which can be used to determine the effects of changing manufacturing parameters and scalability. Static characterization showed substantial local nonlinear behavior that was replicated by new physics-based finite element models, and global linear bending behavior that was modeled using classical shell finite elements incorporating effective properties in place of bulk material properties to represent the unique stiffening structure of these panels. Dynamic characterization was performed on individual panels using standard impact hammer and accelerometer testing, enabling successful extraction of several structural natural frequencies and mode shapes. Additionally, the three dimensional time history of the surface of the panels was rendered from video data, and temporal filters were applied to the data to examine the frequency content. These data were also correlated to the shell element numerical models. Overall, the research contributes to the total knowledge base of gossamer technologies, advances stiff panel-based structures toward space qualification, and demonstrates their potential for use in apertures and other spacecraft.

Métodos tradicionais de animação de fluidos têm dificuldade em resolver escoamentos que envolvem aberturas estreitas e geometrias finas. Abordagens anteriores artificialmente inflaram ou voxelizaram geometrias de objetos finos, sacrificando a geometria e topologias corretas do domínio de simulação, impedindo que o escoamento interaja corretamente com regiões estreitas. No trabalho desenvolvido, apresentamos uma técnica de simulação de fluidos que respeita geometrias complexas de maneira precisa e supera obstáculos comuns em ambientes com aberturas estreitas e geometrias finas. A nossa solução baseia-se no recorte preciso de células do grid regular, gerando uma malha conformal à geometria e topologicamente correta. Nós utilizamos uma abordagem de bordas incorporadas (cut-cells): em cada passo do tempo, a malha de triângulos representando a superfície sólida de um objeto no domínio de simulação é recortada pelas células que intercepta, potencialmente gerando múltiplas sub-células distintas. A malha resultante é conformal ao objeto incorporado e se reduz ao grid regular em regiões que não estão em contato com a superfície. Nós estendemos as abordagens tradicionais de advecção de velocidade e projeção da pressão para dar suporte a essa estrutura de malha aprimorada. Em geral, nossa abordagem é capaz de representar melhor detalhes de geometrias que são menores que uma célula do grid, corretamente recuperando condições de contorno no-slip e free-slip, enquanto mantém uma convergência para a solução da pressão de segunda ordem no espaço. Para melhorar a advecção em regiões próximas às bordas irregulares, introduzimos um método de interpolação que funciona em células poliédricas arbitrárias, utilizando-se do método de interpolação spherical barycentric coordinates (SBC). Essa abordagem possibilita que as linhas características do escoamento respeitem a geometria sem penetrá-la, em contraste com métodos tradicionais de interpolação lineares ou cúbicos. Finalmente, nós melhoramos os métodos de advecção com um método FLIP modificado. Nosso método resolve uma dificuldade inerente a advecção Semi-Lagrangiana no contexto de geometrias deslocando-se através do domínio de simulação: as células que são varridas por sólidos em locomoção perdem sua informação de velocidade e tem de ser preenchidas com velocidades extrapoladas de células vizinhas. Nosso esquema FLIP garante que a informação de velocidade viaje corretamente com as superfícies, não necessitando de nenhum método de extrapolação. / Fluid animation methods based on Eulerian grids have long struggled to resolve flows involving narrow gaps and thin solid features. Past approaches have artificially inflated or voxelized boundaries, although this sacrifices the correct geometry and topology of the fluid domain and prevents flow through narrow regions. We present a boundary-respecting fluid simulator that overcomes these challenges. Our solution is to intersect the solid boundary geometry with the cells of a background regular grid to generate a topologically correct, boundary-conforming cut-cell mesh. We extend both pressure projection and velocity advection to support this enhanced grid structure. For pressure projection, we introduce a general graph-based scheme that properly preserves discrete incompressibility even in thin and topologically complex flow regions, while nevertheless yielding symmetric positive definite linear systems. For advection, we exploit polyhedral interpolation to improve the degree to which the flow conforms to irregular and possibly non-convex cell boundaries, and propose a modified PIC/FLIP advection scheme to eliminate the need to inaccurately reinitialize invalid cells that are swept over by moving boundaries. The method naturally extends the standard Eulerian fluid simulation framework, and while we focus on thin boundaries, our contributions are beneficial for volumetric solids as well. Our results demonstrate successful one-way fluid-solid coupling in the presence of thin objects and narrow flow regions even on very coarse grids.

An Eulerian approach to fluid flow provides an efficient, stable paradigm for realistic fluid simulation. However, its traditional reliance on a fixed-resolution grid is not ideal for simulations that simultaneously exhibit both large and small-scale fluid phenomena. Octree-based fluid simulation approaches have provided the needed adaptivity, but the inherent weakness of a pointer-based tree structure has limited their effectiveness. We present a linear octree structure that provides a significant runtime speedup using these octree-based simulation algorithms. As memory prices continue to decline, we leverage additional memory when compared to traditional octree structures to provide this improvement. In addition to reducing the level of indirection in the data, because our linear octree is stored contiguously in memory as a simple C array rather than a recursive set of pointers, we provide a more cache-friendly data layout than a traditional octree. In our testing, our approach yielded run-times that were 1.5 to nearly 5 times faster than the same simulations running on a traditional octree implementation.