O objetivo deste trabalho é desenvolver um sistema de navegação global para que veículos agrícolas autônomos possam executar missões em campos de cultivo através de um sistema de planejamento de trajetórias. Missões podem ser entendidas como sendo tarefas (p.ex.: de monitoramento, coleta de amostras, etc.) através de pequenas rotas que os veículos devem seguir ao longo de seus trabalhos diários, percorrendo a menor distância possível entre os pontos de origem e destino. O planejamento de trajetória foi dividido em etapas para facilitar o entendimento de cada uma delas. O mapeamento apresentado neste trabalho foi feito em regiões de cultivo de café nos estados de São Paulo e Minas Gerais. Os pontos do mapa foram amostrados utilizando um módulo receptor de sinal GPS (Global Positioning System) ao longo dos caminhos onde é possível a passagem do veículo dentro da plantação. Uma etapa importante para o sucesso deste sistema é a etapa de pré-processamento dos dados. Nesta etapa são inseridas as relações entre os pontos do mapeamento da área. As missões foram pré-definidas de modo a testar o cálculo do caminho de custo mínimo que é realizado através do algoritmo de Dijkstra. A cada ponto da rota é fornecido o ângulo de direção com o qual o veículo deve estar em relação ao Norte geográfico. De acordo com a mudança pretendida do ângulo de direção é proposta uma suavização nesta mudança através da alteração do percurso para um arco de circunferência. Neste caso, o raio de giro é informado. A última etapa consiste em fornecer a velocidade máxima de deslocamento do veículo em função da mudança de direção e velocidade angular máxima do centro de massa do veículo. O sistema proposto neste trabalho foi capaz de determinar o caminho com a menor distância entre dois pontos do mapeamento (coordenadas geográficas) e o calcular da distância entre os pontos. Embora a fórmula utilizada para calcular a distância entre duas coordenadas geográficas considerar o formato da Terra como sendo uma esfera, isto não gerou erro significativo para a aplicação proposta. A suavização proposta possibilitou, em alguns pontos, o aumento da velocidade de deslocamento por fazer a mudança do ângulo de direção de forma menos abrupta. / The objective of this work is to develop a global navigation system for autonomous agricultural vehicles can perform missions in crop fields through a system of path planning. Missions can be understood as tasks (eg monitoring, sampling, etc.). Through small routes that vehicles must follow throughout their daily jobs, traveling the shortest possible distance between the points of origin and destination. The path planning was divided into steps to make it easy to understand each one. The mapping presented in this work was done in coffee-growing regions in the state of São Paulo and Minas Gerais. The map points have been sampled using a GPS receiver module along the path where it is possible to move the vehicle within the plantation. An important step for the success of this system is the data pre-processing step. In this step are inserted the relations between the points of the mapping. The missions are predefined in order to test if the calculation of the minimum cost path made by Dijkstra algorithm is correct. At each point of the route is given the vehicle heading angle (vehicle position towards the geographic North). According to the intended change of the heading angle is proposed a smoothing method to smooth this change by changing the route to an arc. In this case, the turning radius is reported. The last step is to provide the maximum speed of the vehicle due to the change of direction and maximum angular speed of the center of mass. The system proposed in this paper was able to determine the path with the shortest distance between two points of the mapping (geographic coordinates) and calculate the distance between these points. Although the formula used to calculate the distance between two geographical coordinates consider the shape of the Earth as a sphere, this did not generate significant errors for the proposed application. The proposed smoothing allowed, in some cases, to increase the vehicle speed by making the change of heading angle less abrupt.

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-07-25T17:21:34Z No. of bitstreams: 1 RaphaelSantos.pdf: 3718930 bytes, checksum: df335fd5562e8156000972c282fe9724 (MD5) / Made available in DSpace on 2017-07-25T17:21:34Z (GMT). No. of bitstreams: 1 RaphaelSantos.pdf: 3718930 bytes, checksum: df335fd5562e8156000972c282fe9724 (MD5) Previous issue date: 2016-08-17 / Autonomous exploration, in robotics, can be defined as the act of moving into an unknown environment, at priori, while building up a map of the environment. A great deal of literature describes several problems that are relate to the strategy exploration: perception, location, trajectory control and mapping. This work aims to present an autonomous exploration algorithm based on metaheuristics. Therefore, the problem of autonomous exploration of mobile robots is formulated as an optimization problem, providing data for metaheuristics that are able to search points in the space of solutions that represent positions on the map under construction that best meet the objectives of the exploration. Metaheuristics are approximate methods that guarantee sufficiently good solutions to optimization problems. The proposal was implemented and incorporated as an optimization module in a simultaneous location and mapping system that was run on the Robot Operating System environment and proved to be able to guide a simulated robot without human intervention. Two optimization metaheuristics were implemented to guide target to simulated robot: Genetic Algorithm and Firefly Algorithm. Both algorithms have achieved good results, however the second one was able to guide robot by best trajectories. / Exploração autônoma, em robótica, pode ser definida como o ato de mover-se em um ambiente, a princípio desconhecido, enquanto constrói-se um mapa deste ambiente. Uma grande parte da literatura relata vários problemas que se relacionam com a estratégia de exploração: percepção, localização, trajetória, controle e mapeamento. Este trabalho visa apresentar um algoritmo de exploração autonoma baseado em metaheurísticas. Para tanto, o problema de exploração autônoma de robôs móveis é formulado como um problema de otimização, fornecendo dados para que metaheurísticas sejam capazes de buscar pontos no espaço de soluções que representam posições no mapa em construção que melhor satisfaçam os objetivos da exploração. Metaheuristicas são metodos aproximados que garantem soluções suficientemente boas para problemas de otimização. A proposta foi implementada e incorporada como um módulo de otimização em um sistema de localização e mapeamento simultâneos que foi executado em ambiente Robot Operating System e mostrou-se capaz de guiar um robô simulado sem intervenção humana. As metaheurísticas usadas foram o Algoritmo Genético e o Algoritmo de Vagalumes. Ambos os algoritmos obtiveram bons resultados, no entanto o Algoritmo de Vagalumes guiou o robô por trajetórias melhores.

Aquatic robots, such as Autonomous Underwater Vehicles (AUVs), play a major role in the study of ocean processes that require long-term sampling efforts and commonly perform navigation via dead-reckoning using an accelerometer, a magnetometer, a compass, an IMU and a depth sensor for feedback. However, these instruments are subjected to large drift, leading to unbounded uncertainty in location. Moreover, the spatio-temporal dynamics of the ocean environment, coupled with limited communication capabilities, make navigation and localization difficult, especially in coastal regions where the majority of interesting phenomena occur. To add to this, the interesting features are themselves spatio-temporally dynamic, and effective sampling requires a good understanding of vehicle localization relative to the sampled feature. Therefore, our work is motivated by the desire to enable intelligent data collection of complex dynamics and processes that occur in coastal ocean environments to further our understanding and prediction capabilities. The study originated from the need to localize and navigate aquatic robots in a GPS-denied environment and examine the role of the spatio-temporal dynamics of the ocean into the localization and navigation processes. The methods and techniques needed range from the data collection to the localization and navigation algorithms used on-board of the aquatic vehicles. The focus of this work is to develop algorithms for localization and navigation of AUVs in GPS-denied environments. We developed an Augmented terrain-based framework that incorporates physical science data, i.e., temperature, salinity, pH, etc., to enhance the topographic map that the vehicle uses to navigate. In this navigation scheme, the bathymetric data are combined with the physical science data to enrich the uniqueness of the underlying terrain map and increase the accuracy of underwater localization. Another technique developed in this work addresses the problem of tracking an underwater vehicle when the GPS signal suddenly becomes unavailable. The methods include the whitening of the data to reveal the true statistical distance between datapoints and also incorporates physical science data to enhance the topographic map. Simulations were performed at Lake Nighthorse, Colorado, USA, between April 25th and May 2nd 2018 and at Big Fisherman's Cove, Santa Catalina Island, California, USA, on July 13th and July 14th 2016. Different missions were executed on different environments (snow, rain and the presence of plumes). Results showed that these two methodologies for localization and tracking work for reference maps that had been recorded within a week and the accuracy on the average error in localization can be compared to the errors found when using GPS if the time in which the observations were taken are the same period of the day (morning, afternoon or night). The whitening of the data had positive results when compared to localizing without whitening.

This report describes a docking system to allow autonomous battery charging of a mobile robot, and a Web interface that allows long-term unaided use of a sophisticated mobile robot by untrained Web users around the world. The docking system and Web interface are applied to the biologically inspired RatSLAM system as a foundation for testing both its long-term stability and its practicality for real-world applications. While there are existing battery charging and Web interface systems for mobile robots, the developed solution combines the two, resulting in a self-sufficient robot that can recharge its own batteries and stay accessible from the Web. Existing mobile robots on the Internet require manual charging by a human operator, leading to significant periods when the robot is offline. Furthermore, since the robot may be operational for extended periods without powering down, it may perform learning operations that require significantly longer operation than a single battery-recharge cycle would allow. The implemented Web interface makes use of the RatSLAM navigation system. RatSLAM provides the onboard intelligence for the robot to navigate to the user-supplied goal locations (such as “go to location X”), despite long paths or obstacles in the environment. The majority of the existing robot interfaces on the Internet provide direct control of the robot (such as “drive forward”) and therefore the users suffer greatly from the inherent delays of the Internet due to the time lag between performing an action and seeing the feedback. Instead, the robot in this study uses an onboard intelligent navigation system to generate all low-level commands. Due to the minimal input required to give high-level commands to the robot, the system is robust to the long and highly unpredictable delays of Internet communication. Traditional methods of autonomous battery charging for mobile robots have had limited reliability, often due to the mechanical docking system requiring a highly precise connection. Therefore, the mechanical design of the implemented battery charging system improves reliability by allowing for a significantly larger navigation error. In addition, the robot uses a standard vision sensor for both the long-range and short-range stages of navigation to the battery charger, compared to the many systems that require an omnidirectional camera and a high-resolution Laser range finder for this process. The result of this study is a public web interface at "http://ratslam.itee.uq.edu.au/robot.html" (currently offline), where any Web user in the world can watch and control the live mobile robot that is using RatSLAM for navigation, as it drives in its laboratory environment without human assistance.

Navigation in mobile robotics involves two tasks, keeping track of the robot's position and moving according to a control strategy. In addition, when no prior knowledge of the environment is available, the problem is even more difficult, as the robot has to build a map of its surroundings as it moves. These three problems ought to be solved in conjunction since they depend on each other. This thesis is about simultaneously controlling an autonomous vehicle, estimating its location and building the map of the environment. The main objective is to analyse the problem from a control theoretical perspective based on the EKF-SLAM implementation. The contribution of this thesis is the analysis of system's properties such as observability, controllability and stability, which allow us to propose an appropriate navigation scheme that produces well-behaved estimators, controllers, and consequently, the system as a whole. We present a steady state analysis of the SLAM problem, identifying the conditions that lead to partial observability. It is shown that the effects of partial observability appear even in the ideal linear Gaussian case. This indicates that linearisation alone is not the only cause of SLAM inconsistency, and that observability must be achieved as a prerequisite to tackling the effects of linearisation. Additionally, full observability is also shown to be necessary during diagonalisation of the covariance matrix, an approach often used to reduce the computational complexity of the SLAM algorithm, and which leads to full controllability as we show in this work.Focusing specifically on the case of a system with a single monocular camera, we present an observability analysis using the nullspace basis of the stripped observability matrix. The aim is to get a better understanding of the well known intuitive behaviour of this type of systems, such as the need for triangulation to features from different positions in order to get accurate relative pose estimates between vehicle and camera. Through characterisation the unobservable directions in monocular SLAM, we are able to identify the vehicle motions required to maximise the number of observable states in the system. When closing the control loop of the SLAM system, both the feedback controller and the estimator are shown to be asymptotically stable. Furthermore, we show that the tracking error does not influence the estimation performance of a fully observable system and viceversa, that control is not affected by the estimation. Because of this, a higher level motion strategy is required in order to enhance estimation, specially needed while performing SLAM with a single camera. Considering a real-time application, we propose a control strategy to optimise both the localisation of the vehicle and the feature map by computing the most appropriate control actions or movements. The actions are chosen in order to maximise an information theoretic metric. Simulations and real-time experiments are performed to demonstrate the feasibility of the proposed control strategy.

As control systems are becoming more and more complex, system complexity is rapidly becoming a limiting factor in the efficacy of established techniques for control systems design. To cope with the growing complexity, control architectures often have a hierarchical structure. At the base of the system pyramid lie feedback loops with simple closed-loop control laws. These are followed, at a higher level, by discrete control logics. Such hierarchical systems typically have a hybrid nature. A common approach to addressing these types of complexity consists of decomposing, in the time domain, the control task into a number of modes, i.e. control laws dedicated to carrying out a limited task. This type of control generally involves switching laws among the various modes, and its design poses a major challenge in many application domains. The primary goal of this thesis is to develop a unified framework for addressing this challenge. To this end, the contribution of this thesis is threefold: 1. An algorithmic framework for how to optimize the performance of switched autonomous systems is derived. The optimization concerns both the sequence in which different modes appear in and the duration of each mode. The optimization algorithms are presented together with detailed convergence analyses. 2. Control strategies for how to optimize switched autonomous systems operating in real time, and when the initial state of the system is unknown, are presented. 3. A control strategy for how to optimally navigate an autonomous mobile robot in real-time is presented and evaluated on a mobile robotics platform. The control strategy uses optimal switching surfaces for when to switch between different modes of operations (behaviors).

In this thesis, we first present a contribution to overcome this problem of robustness for the recognition of object instances, then we straightly extend this contribution to the detection and localization of classes of objects. In a first step, we have developed a method inspired by graph matching to address the problem of fast recognition of instances of specific objects in noisy conditions. This method allows to easily combine any types of local features (eg contours, textures ...) less affected by noise than keypoints, while bypassing the normalization problem and without penalizing too much the detection speed. Unlike other methods based on a global rigid transformation, our approach is robust to complex deformations such as those due to perspective or those non-rigid inherent to the model itself (e.g. a face, a flexible magazine). Our experiments on several datasets have showed the relevance of our approach. It is overall slightly less robust to occlusion than existing approaches, but it produces better performances in noisy conditions. In a second step, we have developed an approach for detecting classes of objects in the same spirit as the bag-of-visual-words model. For this we use our cascaded micro-classifiers to recognize visual words more distinctive than the classical words simply based on visual dictionaries. Training is divided into two parts: First, we generate cascades of micro-classifiers for recognizing local parts of the model pictures and then in a second step, we use a classifier to model the decision boundary between images of class and those of non-class. We show that the association of classical visual words (from keypoints patches) and our disctinctive words results in a significant improvement. The computation time is generally quite low, given the structure of the cascades that minimizes the detection time and the form of the classifier is extremely fast to evaluate.

This thesis examines the use of multiple robots in cooperative simultaneous localization and mapping (SLAM), where each robot must estimate the poses of all robots in the team, along with the positions of all known landmarks. The robot team must operate under the condition that the communication network between robots is never guaranteed to be fully connected. Under this condition, a novel algorithm is derived that allows each robot to obtain the centralized-equivalent estimate in a decentralized manner, whenever possible. The algorithm is then extended to a decentralized and distributed approach where robots share the computational burden in considering different data association hypotheses in generating the centralized-equivalent consensus estimate.

A mobile robotic platform for Autonomous Site Selection and Preparation (ASSP) was developed for an analogue deployment to Mauna Kea, Hawai‘i. A team of rovers performed an autonomous Ground Penetrating Radar (GPR) survey and constructed a level landing pad. They used interchangeable payloads that allowed the GPR and blade to be easily exchanged. Autonomy was accomplished by integrating the individual hardware devices with software based on the ArgoSoft framework previously developed at UTIAS. The rovers were controlled by an on-board netbook. The successes and failures of the devices and software modules are evaluated within. Recommendations are presented to address problems discovered during the deployment and to guide future research on the platform.

This thesis examines the use of multiple robots in cooperative simultaneous localization and mapping (SLAM), where each robot must estimate the poses of all robots in the team, along with the positions of all known landmarks. The robot team must operate under the condition that the communication network between robots is never guaranteed to be fully connected. Under this condition, a novel algorithm is derived that allows each robot to obtain the centralized-equivalent estimate in a decentralized manner, whenever possible. The algorithm is then extended to a decentralized and distributed approach where robots share the computational burden in considering different data association hypotheses in generating the centralized-equivalent consensus estimate.