At first glance, robots and proteins have little in common. Robots are commonly thought of as tools that perform tasks such as vacuuming the floor, while proteins play essential roles in many biochemical processes. However, the functionality of both robots and proteins is highly dependent on their motions. In order to study motions in these two divergent domains, the same underlying algorithmic framework can be applied. This method is derived from probabilistic roadmap methods (PRMs) originally developed for robotic motion planning. It builds a graph, or roadmap, where configurations are represented as vertices and transitions between configurations are edges. The contribution of this work is a set of intelligent methods applied to PRMs. These methods facilitate both the modeling and analysis of motions, and have enabled the study of complex and high-dimensional problems in both robotic and molecular domains. In order to efficiently study biologically relevant molecular folding behaviors we have developed new techniques based on Monte Carlo solution, master equation calculation, and non-linear dimensionality reduction to run simulations and analysis on the roadmap. The first method, Map-based master equation calculation (MME), extracts global properties of the folding landscape such as global folding rates. On the other hand, another method, Map-based Monte Carlo solution (MMC), can be used to extract microscopic features of the folding process. Also, the application of dimensionality reduction returns a lower-dimensional representation that still retains the principal features while facilitating both modeling and analysis of motion landscapes. A key contribution of our methods is the flexibility to study larger and more complex structures, e.g., 372 residue Alpha-1 antitrypsin and 200 nucleotide ColE1 RNAII. We also applied intelligent roadmap-based techniques to the area of robotic motion. These methods take advantage of unsupervised learning methods at all stages of the planning process and produces solutions in complex spaces with little cost and less manual intervention compared to other adaptive methods. Our results show that our methods have low overhead and that they out-perform two existing adaptive methods in all complex cases studied.

As technology use increases in the world and inundates everyday life, the visual aspect of technology or computer graphics becomes increasingly important. This thesis presents a system for the automatic generation of virtual camera paths for fly-throughs of a digital scene. The sample scene used in this work is an underwater setting featuring a shipwreck model with other virtual underwater elements such as rocks, bubbles and caustics. The digital shipwreck model was reconstructed from an actual World War II shipwreck, resting off the coast of Malta. Video and sonar scans from an autonomous underwater vehicle were used in a photogrammetry pipeline to create the model. This thesis presents an algorithm to automatically generate virtual camera paths using a robotics motion planning algorithm, specifically the probabilistic roadmap. This algorithm uses a rapidly-exploring random tree to quickly cover a space and generate small maps with good coverage. For this work, the camera pitch and height along a specified path were automatically generated using cinematographic and geometric principles. These principles were used to evaluate potential viewpoints and influence whether or not a view is used in the final path. A computational evaluation of ‘the rule of thirds’ and evaluation of the model normals relative to the camera viewpoint are used to represent cinematography and geometry principles. In addition to the system that automatically generates virtual camera paths, a user study is presented which evaluates ten different videos produced via camera paths with this system. The videos were created using different viewpoint evaluation methods and different path generation characteristics. The user study indicates that users prefer paths generated by our system over flat and randomly generated paths. Specifically, users prefer paths generated using the computational evaluation of the rule of thirds and paths that show the wreck from a large variety of angles but without too much camera undulation.

O planejamento probabilístico de mapa de rotas tem se mostrado uma poderosa ferramenta para o planejamento de caminhos para robôs móveis, devido a sua eficiência computacional, simplicidade de implementação e escalabilidade em diferentes problemas. Este método de planejamento possui duas fases. Na fase de construção, um mapa de rotas é gerado de forma iterativa e incremental, e armazenado na forma de um grafo G, cujos vértices são configurações livres, amostradas no espaço de configurações do robô e cujas arestas correspondem a caminhos livres de colisão entre tais configurações. Na fase de questionamento, dadas quaisquer configurações de origem e destino, \'alfa\' e \'beta\' respectivamente, o planejador conecta \'alfa\' e \'beta\' à G inserindo arestas que correspondem a caminhos livres de colisão, para então procurar por um caminho entre \'alfa\' e \'beta\' em G. Neste trabalho o foco reside principalmente na fase de construção do mapa de rotas. O objetivo aqui consiste em efetuar uma análise comparativa de diversas combinações de diferentes técnicas de amostragem das configurações livres e de diferentes técnicas de seleção de vértices em G, todas implementadas em um único sistema e aplicadas aos mesmos cenários. Os resultados propiciam um valioso auxílio aos usuários do planejamento probabilístico de mapas de rotas na decisão da melhor combinação para suas aplicações. / The probabilistic roadmap planning has emerged as a powerful framework for path planning of mobile robots due to its computational efficiency, implementation simplicity, and scalability in different problems. This planning method proceeds in two phases. In the construction phase a roadmap is incrementally constructed and stored as a graph G whose nodes are free configurations sampled on the robot\'s configuration space and whose edges correspond to collision-free paths between these configurations. In the query phase, given any start and goal configurations, \'alfa\' and \'beta\' respectively, the planner first connects \'alfa\' and \'beta\' to G by adding edges that correspond to collision-free paths, and then searches for a path in G between \'alfa\' and \'beta\'. In this work, we address mainly the roadmap construction phase. The goal here is to provide a comparative analysis of a number of combinations of different techniques for sampling free configurations and different node adding techniques, all implemented in a single system and applied to the same test workspace. Results help probabilistic roadmap planning users to choose the best combination for their applications.

The visibility-aware path planner addresses the problem of path planning for target visibility. It computes sequences of motions that afford a line of sight to a stationary visual target for sensors on a robotic platform. The visibility-aware planner uses a model of the visible region, namely, the region of the task space in which a line of sight exists to the target. The planner also takes the orientation of the sensor into account, utilizing a model of the field of view frustum. The planner applies a penalty to paths that cause the sensor to lose target visibility by exiting the visible region or rotating so the target is not in the field of view. The planner applies these penalties to the edges in a probabilistic roadmap, providing weights in the roadmap graph for graph-search based planning algorithms. This thesis presents two variants on the planner. The static multi-query planner precomputes penalties for all roadmap edges and performs a best-path search using Dijkstra's algorithm. The dynamic single-query planner uses an iterative test-and-reject search to find paths of acceptable penalty without the benefit of precomputation. Four experiments are presented which validate the planners and present examples of the path planning for visibility on 6-DOF robot manipulators. The algorithms are statistically tested with multiple queries. Results show that the planner finds paths with significantly lower losses of target visibility than existing shortest-path planners.

Protein folding is considered to be one of the grand challenge problems in biology. Protein folding refers to how a protein's amino acid sequence, under certain physiological conditions, folds into a stable close-packed three-dimensional structure known as the native state. There are two major problems in protein folding. One, usually called protein structure prediction, is to predict the structure of the protein's native state given only the amino acid sequence. Another important and strongly related problem, often called protein folding, is to study how the amino acid sequence dynamically transitions from an unstructured state to the native state. In this dissertation, we concentrate on the second problem. There are several approaches that have been applied to the protein folding problem, including molecular dynamics, Monte Carlo methods, statistical mechanical models, and lattice models. However, most of these approaches suffer from either overly-detailed simulations, requiring impractical computation times, or overly-simplified models, resulting in unrealistic solutions. In this work, we present a novel motion planning based framework for studying protein folding. We describe how it can be used to approximately map a protein's energy landscape, and then discuss how to find approximate folding pathways and kinetics on this approximate energy landscape. In particular, our technique can produce potential energy landscapes, free energy landscapes, and many folding pathways all from a single roadmap. The roadmap can be computed in a few hours on a desktop PC using a coarse potential energy function. In addition, our motion planning based approach is the first simulation method that enables the study of protein folding kinetics at a level of detail that is appropriate (i.e., not too detailed or too coarse) for capturing possible 2-state and 3-state folding kinetics that may coexist in one protein. Indeed, the unique ability of our method to produce large sets of unrelated folding pathways may potentially provide crucial insight into some aspects of folding kinetics that are not available to other theoretical techniques.

The visibility-aware path planner addresses the problem of path planning for target visibility. It computes sequences of motions that afford a line of sight to a stationary visual target for sensors on a robotic platform. The visibility-aware planner uses a model of the visible region, namely, the region of the task space in which a line of sight exists to the target. The planner also takes the orientation of the sensor into account, utilizing a model of the field of view frustum. The planner applies a penalty to paths that cause the sensor to lose target visibility by exiting the visible region or rotating so the target is not in the field of view. The planner applies these penalties to the edges in a probabilistic roadmap, providing weights in the roadmap graph for graph-search based planning algorithms. This thesis presents two variants on the planner. The static multi-query planner precomputes penalties for all roadmap edges and performs a best-path search using Dijkstra's algorithm. The dynamic single-query planner uses an iterative test-and-reject search to find paths of acceptable penalty without the benefit of precomputation. Four experiments are presented which validate the planners and present examples of the path planning for visibility on 6-DOF robot manipulators. The algorithms are statistically tested with multiple queries. Results show that the planner finds paths with significantly lower losses of target visibility than existing shortest-path planners.

O planejamento probabilístico de mapa de rotas tem se mostrado uma poderosa ferramenta para o planejamento de caminhos para robôs móveis, devido a sua eficiência computacional, simplicidade de implementação e escalabilidade em diferentes problemas. Este método de planejamento possui duas fases. Na fase de construção, um mapa de rotas é gerado de forma iterativa e incremental, e armazenado na forma de um grafo G, cujos vértices são configurações livres, amostradas no espaço de configurações do robô e cujas arestas correspondem a caminhos livres de colisão entre tais configurações. Na fase de questionamento, dadas quaisquer configurações de origem e destino, \'alfa\' e \'beta\' respectivamente, o planejador conecta \'alfa\' e \'beta\' à G inserindo arestas que correspondem a caminhos livres de colisão, para então procurar por um caminho entre \'alfa\' e \'beta\' em G. Neste trabalho o foco reside principalmente na fase de construção do mapa de rotas. O objetivo aqui consiste em efetuar uma análise comparativa de diversas combinações de diferentes técnicas de amostragem das configurações livres e de diferentes técnicas de seleção de vértices em G, todas implementadas em um único sistema e aplicadas aos mesmos cenários. Os resultados propiciam um valioso auxílio aos usuários do planejamento probabilístico de mapas de rotas na decisão da melhor combinação para suas aplicações. / The probabilistic roadmap planning has emerged as a powerful framework for path planning of mobile robots due to its computational efficiency, implementation simplicity, and scalability in different problems. This planning method proceeds in two phases. In the construction phase a roadmap is incrementally constructed and stored as a graph G whose nodes are free configurations sampled on the robot\'s configuration space and whose edges correspond to collision-free paths between these configurations. In the query phase, given any start and goal configurations, \'alfa\' and \'beta\' respectively, the planner first connects \'alfa\' and \'beta\' to G by adding edges that correspond to collision-free paths, and then searches for a path in G between \'alfa\' and \'beta\'. In this work, we address mainly the roadmap construction phase. The goal here is to provide a comparative analysis of a number of combinations of different techniques for sampling free configurations and different node adding techniques, all implemented in a single system and applied to the same test workspace. Results help probabilistic roadmap planning users to choose the best combination for their applications.

The visibility-aware path planner addresses the problem of path planning for target visibility. It computes sequences of motions that afford a line of sight to a stationary visual target for sensors on a robotic platform. The visibility-aware planner uses a model of the visible region, namely, the region of the task space in which a line of sight exists to the target. The planner also takes the orientation of the sensor into account, utilizing a model of the field of view frustum. The planner applies a penalty to paths that cause the sensor to lose target visibility by exiting the visible region or rotating so the target is not in the field of view. The planner applies these penalties to the edges in a probabilistic roadmap, providing weights in the roadmap graph for graph-search based planning algorithms. This thesis presents two variants on the planner. The static multi-query planner precomputes penalties for all roadmap edges and performs a best-path search using Dijkstra's algorithm. The dynamic single-query planner uses an iterative test-and-reject search to find paths of acceptable penalty without the benefit of precomputation. Four experiments are presented which validate the planners and present examples of the path planning for visibility on 6-DOF robot manipulators. The algorithms are statistically tested with multiple queries. Results show that the planner finds paths with significantly lower losses of target visibility than existing shortest-path planners. / Science, Faculty of / Computer Science, Department of / Graduate