O desenvolvimento de arquiteturas de controle para múltiplos robôs em ambientes dinâmicos tem sido tema de pesquisas na área de robótica. A complexidade deste tema varia de acordo com as necessidades exigidas da equipe de robôs. Em geral, espera-se que os robôs colaborem uns com os outros na execução de uma tarefa. Além disso, cada robô deve ser capaz de planejar trajetórias e replanejá-las em caso de situações inesperadas. No presente trabalho, propomos uma Arquitetura de Controle Inteligente para múltiplos robôs denominada ACIn. Para esta finalidade, foram investigadas algumas técnicas utilizadas para o controle inteligente de robôs, tais como, Redes Neurais Artificiais, Campos Potenciais e Campos Potenciais baseados em Problema do Valor de Contorno (PVC). Tais técnicas, normalmente utilizadas para um único robô, foram adaptadas para tornar possível o controle de múltiplos robôs sob arquitetura ACIn. Uma outra contribuição deste trabalho refere-se ao aperfeiçoamento da técnica de Campos Potenciais baseada PVC denominada Campos Potenciais Localmente Orientados (CPLO). Este aperfeiçoamento foi proposto para suprir a deficiência das técnicas baseadas em PVC quando estas são aplicadas em ambientes com múltiplos robôs. Além disso, um Sistema Baseado em Regras (SBR) também foi proposto como parte integrante da arquitetura ACIn. O objetivo do SBR é caracterizar a funcionalidade de cada robô para uma determinada tarefa. Isto se faz necessário para que o comportamento dos integrantes da equipe de robôs não seja competitivo e sim colaborativo. Por fim, através dos experimentos utilizando o simulador oficial de futebol de robôs da FIRA, observou-se que a arquitetura de controle inteligente (ACIn) implementada com a técnica de planejamento CPLO e SBR propostos, mostrou-se robusta e eficiente no controle de múltiplos robôs / In this work, an intelligent control architecture for multi-robots denominated ACIn was proposed. With this objective, some techniques considered intelligent were studied for the planning of trajectories, such as Artificial Neural Networks, Potential Fields and Potential Fields based on the boundary value problem (BVP). Such techniques, normally used for a single robot, were adapted to function with multi-robots inside the ACIn architecture. Another contribution of this work refers to the improvement of the Potential Fields based on the boundary value problem (BVP) technique. This improvement was proposed to supply the drawback of the BVP based techniques when they are applied to multi robots environments. Besides, a Rule Based System (RBS) was also proposed as part of the ACIn architecture. The objective of the RBS is to characterize the functionality of each robot for a determined task. This is necessary for the behavior of the equip members not to be competitive, but collaborative. Finally, it was observed through the experiments with the robot soccer simulated environment, that our intelligent control architecture (ACIn) proposal, integrating planning and RBS for the control of multi-robots was satisfactory

<p>With the advancements in acquisition and processing of seismic reflection data recorded over crystalline rocks, building three-dimensional geologic models becomes increasingly favorable. Because of little available petrophysical data, interpretations of seismic reflection data in hardrock terrains are often speculative. Potential field data modeling are sometimes performed in order to reduce the ambiguity of seismic reflection interpretations. The Kristineberg mining area in the western part of the Paleoproterozoic Skellefte Ore District was chosen to construct a pilot three-dimensional geologic model in an attempt to understand the crustal architecture in the region and how the major mineral systems operated in this architecture. To contribute to this aim, two parallel seismic reflection profiles were acquired in 2003 and processed to 20 sec with special attention to the top 4 sec of data. Several reflections were imaged and interpreted by the aid of reflector modeling, borehole data, 2.5D and 3D potential field modeling, and geological observations. Interpretations are informative at the crustal scale and help to construct a three-dimensional geologic model of the Kristineberg mining area. The three-dimensional geologic model covers an area of 30×30 km² down to a depth of 12 km. The integrations help to interpret a structural basement to the Skellefte volcanic rocks, possibly with Bothnian Basin metasedimentary affinity. The contact is a shear-zone that separates the two units, generating large fold structures, which can be observed in the region. The interpretations help to divide the Revsund granitic rocks into two major groups based on their present shape and thickness. A large gravity low in the south is best represented by the intrusion of thick dome of Revsund granite. In the north, the low-gravity corresponds to the intrusion of sheet-like Revsund granites. In general, the structure associated with the Skellefte volcanics and the overlying metasedimentary rocks are two thrusts exposing the Skellefte volcanic rocks in the cores of hanging wall anticlinal structures. Lack of coherent reflectivity in the seismic reflection data may be due to complex faulting and folding systems observed in the Skellefte volcanics. Ultramafic sills within the metasedimentary rocks are interpreted to extend down to depths of about 5-6 km. The interpretations are helpful for targeting new VHMS deposits and areas with gold potential. For VHMS deposits, these are situated in the southern limb of a local synformal structure south of the Kristineberg mine, on the contact between the Revsund granite and the Skellefte volcanic rocks. A combination of metasedimentary and mafic-ultramafic rocks are highly gold prospective in the west, similar to observations elsewhere in the region. There are still questions that remain unanswered and need more work. New data in the study area will help to answer questions related to e.g., an enigmatic diffraction seismic signal in Profile 5 and the structural relationship between the Skellefte volcanic rocks and the Malå volcanics. Although the derived 3D geologic model is preliminary and constructed at the crustal scale, it provides useful information to better understand the tectonic evolution of the Kristineberg mining area. </p>

Mobile robots are increasingly being used to perform tasks in unknown environments. The potential of robots to undertake such tasks lies in their ability to intelligently and efficiently locate and interact with objects in their environment. My research focuses on developing algorithms to plan paths for mobile robots in a partially known environment observed by an overhead camera. The environment consists of dynamic obstacles and targets. A new methodology, Extrapolated Artificial Potential Field, is proposed for real time robot path planning. An algorithm for probabilistic collision detection and avoidance is used to enhance the planner. The aim of the robot is to select avoidance maneuvers to avoid the dynamic obstacles. The navigation of a mobile robot in a real-world dynamic environment is a complex and daunting task. Consider the case of a mobile robot working in an office environment. It has to avoid the static obstacles such as desks, chairs and cupboards and it also has to consider dynamic obstacles such as humans. In the presence of dynamic obstacles, the robot has to predict the motion of the obstacles. Humans inherently have an intuitive motion prediction scheme when planning a path in a crowded environment. A technique has been developed which predicts the possible future positions of obstacles. This technique coupled with the generalized Voronoi diagram enables the robot to safely navigate in a given environment.

In this thesis, four inversion problems of different scale and difficulty are solved. Two of them are electromagnetic inverse problems. Two more are joint inversion problems of potential field data and other types of data. First, a linear approximation, which is a generalization of the low-induction-number approximation standard in slingram dual-loop interpretation is developed and used for rapid two and three dimensional inversion. The approximation takes induction within a background half-space into account and can thus be applied in conductive scenarios, where otherwise a rigorous electromagnetic modeling would be required. Second, a three-dimensional inversion of airborne tensor very-low-frequency data with a rigorous forward modeling at its core is developed. For dealing with the large scale of the forward problem, a nested fast-Fourier-transform-based integral equation method is introduced, wherein electromagnetic interactions are arranged according to their range and larger ranges are treated with less accuracy and effort. The inversion improves the traditional interpretation through data derived maps by providing a conductivity model, thus constraining the upper few hundred meters of the crust down to the shallowest conductor and allowing the study of its top in three dimensions. The third inversion problem is the the joint inversion of refraction and geoelectric data. By requiring the velocity and resistivity models to share a common, laterally variable layered geometry, easily interpretable models, which are reasonable in many geological near surface situations (e.g., groundwater exploration in Quaternary sediments), are produced directly from the joint inversion. Finally, a joint inversion of large scale potential field data from a gabbro intrusion is presented. Gravity and magnetic data are required to abide to a petrophysical constraint, which is derived from extensive field sampling. The impact of the constraint is maximized under the provision that both data sets are explained equally well as they would be through individual inversions. This leads to a simple and clearly defined intrusion geometry, consistent for both the density and magnetic susceptibility distribution. In all presented inversion problems, field data sets are successfully inverted, the results are appraised through synthetic tests and, if available, through comparison with independent data. / Diese Arbeit hat die Lösung von vier geophysikalischen Umkehraufgaben, sogenannten Inversionsproblemen, zum Gegenstand. Zwei dieser Aufgaben befassen sich mit der Inversion elektromagnetischer Daten, zwei weitere sind Probleme der kombinierten Inversion von Datensätzen aus unterschiedlichen geophysikalischen Messverfahren. Im ersten Problem wird die für die Auswertung elektromagnetischer Zweispulensystemdaten typische lineare Näherung der kleinen Induktionszahlen als Bornsche Näherung verallgemeinert, ihre Anwendbarkeit durch exakte Berücksichtigung der Induktionsvorgänge in einem beliebigen homogenen Halbraum von schlechtleitenden auf gutleitende Untergründe ausgedehnt und schließlich zur zwei- und dreidimensionalen Inversion eingesetzt. Dadurch kann auch im leitfähigen Untergrund eine aufwändige exakte Modellierung vermieden werden. Im zweiten Problem wird eine dreidimensionale Inversion von flugzeuggestützten Längstwellenmessungen entwickelt und als ihre Grundlage eine exakte elektromagnetische Rechnung erdacht. Damit wird traditionelle kartengestützte Dateninterpretation durch ein dreidimensionales Leitfähigkeitsmodell ergänzt, welches die oberen hundert bis dreihundert Meter der Erdkruste bis hin zur Tiefe des obersten Leiters abbildet, so dass dessen Oberflächenform erkundet werden kann. Die enorme Problemgröße wird durch eine Fouriertransformationsmethode bewältigt, welche die elektromagnetischen Wechselwirkungen nach ihrer Reichweite einteilt, die Fernwirkungen mit entsprechend verringerter Genauigkeit behandelt und dadurch eine erhebliche Anzahl an Rechnungen einspart. Im dritten Problem werden refraktionsseismische und geoelektrische Messungen kombiniert, indem sowohl das Geschwindigkeits- als auch das Widerstandsmodell mit einer gemeinsamen, lateral veränderlichen und durch beide Datensätze bestimmten Schichtstruktur versehen werden. Ein solches, durch Schichten definiertes Inversionsergebnis, stellt in vielen oberflächennahen Anwendungen, beispielsweise im Grundwasserbereich, ein sinnvolles Abbild der Erde dar. Im vierten Problem werden Schweremessungen und Magnetfeldmessungen, die über einer Gabbrointrusion aufgenommen wurden, mittels einer empirischen petrophysikalischen Beziehung vereinigt, welche aus Labormessungen an einer großen Anzahl von Gesteinsproben abgeleitet wurde. Hierbei wird der Einfluss dieser Modellkopplung solange maximiert, wie beide Datensätze mit derjenigen Genauigkeit angepasst werden können, welche vorher in Einzelinversionen erreicht wurde. Das Ergebnis ist ein einfaches, geometrisch konsistentes Modell der Verteilungen von Dichte und magnetischer Suszeptibilität. In allen vier Aufgaben wurden erfolgreich reale Felddaten invertiert. Die Güte der Ergebnisse wurde mittels synthetischer Experimente untersucht und, so vorhanden, mit unabhängigen Informationen verglichen.

In this work, reinforcement learning algorithms are studied with the help of potential field methods, using robosoccer simulators as test beds. Reinforcement Learning (RL) is a framework for general problem solving where an agent can learn through experience. The soccer game is selected as the problem domain a way of experimenting multi-agent team behaviors because of its popularity and complexity.

Nós refinamos o planejador introduzindo uma nova forma para o núcleo da equação que permite facilmente lidar com terrenos não-homogêneos. Isto é obtido através de mudanças locais na concavidade/convexidade do potencial, criando regiões com altas ou baixas preferências de navegação. Nós integramos esta nova equação ao planejador hierárquico, surgindo uma ampla variedade de aplicações. Nossa proposta contribui para diversas áreas incluindo a navegação de agentes, pathfinding em jogos, simulação de multidões, e a navegação de robôs. Nossas publicações reforçam a relevância e robustez do método proposto. / In this work, we propose a new solution to agent navigation based upon boundary value problems (BVP), called Configurable Flows, to control steering behaviors of characters in dynamic environments. We use a potential field formalism that allows synthetic actors to move negotiating space, avoiding collisions, and attaining goals while producing very individual paths. The individuality of each character can be set by changing its inner field parameters leading to a broad range of possible behaviors without jeopardizing its performance. BVP Path Planners generate potential fields through a differential equation whose gradient descent represents navigation routes from any point of the environment to a goal position. Resulting paths are smooth and free from local minima. In spite of these advantages, these kind of planners consumes a lot of time to produce a solution. Our approach combines a BVP Path Planner with the Full Multigrid Method, which solves elliptic partial differential equations using a hierarchical strategy. The proposed planner enables real-time performance in large environments. Results show that our proposal spends less than 1% of the time needed to compute a solution using the original BVP planners in several environments. We refine our Path Planner by introducing a new form of the core equation that permits to easily cope with terrain inhomogeneities. This is accomplished by locally changing the concavity/ convexity of the potential, and then creating regions with higher or lower navigation preferences. As the potential field requires several steps to converge, this approach can be expensive computationally. To overcome this problem, we integrate this novel core equation to the hierarchical planner, emerging a wide variety of applications. We believe our proposal can contribute to several areas of research including agent navigation, pathfinding for games, crowd simulation and robotics. Our publications reinforce the relevance of the proposed method.

Nós refinamos o planejador introduzindo uma nova forma para o núcleo da equação que permite facilmente lidar com terrenos não-homogêneos. Isto é obtido através de mudanças locais na concavidade/convexidade do potencial, criando regiões com altas ou baixas preferências de navegação. Nós integramos esta nova equação ao planejador hierárquico, surgindo uma ampla variedade de aplicações. Nossa proposta contribui para diversas áreas incluindo a navegação de agentes, pathfinding em jogos, simulação de multidões, e a navegação de robôs. Nossas publicações reforçam a relevância e robustez do método proposto. / In this work, we propose a new solution to agent navigation based upon boundary value problems (BVP), called Configurable Flows, to control steering behaviors of characters in dynamic environments. We use a potential field formalism that allows synthetic actors to move negotiating space, avoiding collisions, and attaining goals while producing very individual paths. The individuality of each character can be set by changing its inner field parameters leading to a broad range of possible behaviors without jeopardizing its performance. BVP Path Planners generate potential fields through a differential equation whose gradient descent represents navigation routes from any point of the environment to a goal position. Resulting paths are smooth and free from local minima. In spite of these advantages, these kind of planners consumes a lot of time to produce a solution. Our approach combines a BVP Path Planner with the Full Multigrid Method, which solves elliptic partial differential equations using a hierarchical strategy. The proposed planner enables real-time performance in large environments. Results show that our proposal spends less than 1% of the time needed to compute a solution using the original BVP planners in several environments. We refine our Path Planner by introducing a new form of the core equation that permits to easily cope with terrain inhomogeneities. This is accomplished by locally changing the concavity/ convexity of the potential, and then creating regions with higher or lower navigation preferences. As the potential field requires several steps to converge, this approach can be expensive computationally. To overcome this problem, we integrate this novel core equation to the hierarchical planner, emerging a wide variety of applications. We believe our proposal can contribute to several areas of research including agent navigation, pathfinding for games, crowd simulation and robotics. Our publications reinforce the relevance of the proposed method.

Nós refinamos o planejador introduzindo uma nova forma para o núcleo da equação que permite facilmente lidar com terrenos não-homogêneos. Isto é obtido através de mudanças locais na concavidade/convexidade do potencial, criando regiões com altas ou baixas preferências de navegação. Nós integramos esta nova equação ao planejador hierárquico, surgindo uma ampla variedade de aplicações. Nossa proposta contribui para diversas áreas incluindo a navegação de agentes, pathfinding em jogos, simulação de multidões, e a navegação de robôs. Nossas publicações reforçam a relevância e robustez do método proposto. / In this work, we propose a new solution to agent navigation based upon boundary value problems (BVP), called Configurable Flows, to control steering behaviors of characters in dynamic environments. We use a potential field formalism that allows synthetic actors to move negotiating space, avoiding collisions, and attaining goals while producing very individual paths. The individuality of each character can be set by changing its inner field parameters leading to a broad range of possible behaviors without jeopardizing its performance. BVP Path Planners generate potential fields through a differential equation whose gradient descent represents navigation routes from any point of the environment to a goal position. Resulting paths are smooth and free from local minima. In spite of these advantages, these kind of planners consumes a lot of time to produce a solution. Our approach combines a BVP Path Planner with the Full Multigrid Method, which solves elliptic partial differential equations using a hierarchical strategy. The proposed planner enables real-time performance in large environments. Results show that our proposal spends less than 1% of the time needed to compute a solution using the original BVP planners in several environments. We refine our Path Planner by introducing a new form of the core equation that permits to easily cope with terrain inhomogeneities. This is accomplished by locally changing the concavity/ convexity of the potential, and then creating regions with higher or lower navigation preferences. As the potential field requires several steps to converge, this approach can be expensive computationally. To overcome this problem, we integrate this novel core equation to the hierarchical planner, emerging a wide variety of applications. We believe our proposal can contribute to several areas of research including agent navigation, pathfinding for games, crowd simulation and robotics. Our publications reinforce the relevance of the proposed method.

In the last decade video games have made great improvements in terms of arti cial intelligence and visuals. Researchers have also made advancements in the arti cial intelligence eld and some of the latest research papers have been exploring potential elds. This report will cover the background of potential eld and examine some improvements that can be made to increase the performance of the algorithm. The basic idea is to increase performance by making a GPGPU(General purpose graphic processing unit) solution for the creation of potential elds. Several GPGPU implementations are presented where focus has lied on optimizing memory access patterns to increase performance. The results of this thesis show that an optimized GPGPU implementation can give up to 18.5x speedup over a CPU implementation.

Shared autonomous vehicles present a significant opportunity to change the way that urban mobility is viewed by society. By providing a shared mobility platform at a cost lower than has previously been obtainable there are significant possibilites to enable a new era of mobility for consumers. This opportunity, however, comes with significant risks in the form of emissions and increased road usage. Understanding how the risks and benefits of shared autonomous vehicles can be balanced is crucial to be able to adequately prepare for their introduction. One of the primary ways to understand the interplay between the risks and benefits of autonomous vehicles is through the use of computer simulations. However, typically simulations must be defined for a specific area and provide results that are not applicable to a wide range of areas and situations. This work presents the development of a framework that can be used to simulate SAV behaviour at any given region of interest. This framework automates the process of generating a directed non-planar graph using data gathered from the OpenStreetMap project. It further provides tools to generate activity based trips that are statistically similar in time and density to provided data that reflects the trips in the simulation area. In the absence of this data, this work has identified the 2009 National Household Travel Survey as an acceptable surrogate for data specific to a region. The framework then provides methods by which the trip origins and destinations are mapped into the directed non-planar graph representation of the area of interest. This mapping is performed using real-world data including business locations and census data. Finally the framework is capable of simulating the activity of SAV in response to the defined trips given a variety of starting conditions and relocation strategies. In addition to the simulation framework this work presents a novel relocation strategy for unoccupied SAV based on the potential field methods that have been used in robotic navigation. This method provides a continously differentiable function that describes the unmet demand in the service area for a network of shared autonomous vehicles. The tunable parameters of the method are explored by using a design of experiments, and optimal values reflecting different scenarios are identified.The method is also evaluated in the context of both and over- and under-supply of vehicles for the given demand. As a result this method has been shown to provide substantial reductions in the wait time for a vehicle to service a trip with a minimal increase in the total distance that is traveled by all vehicles in the network.