For an autonomous mobile robot, localization and map building are vital capabilities. The localization ability provides the robot location information, so the robot can navigate in the environment. On the other hand, the robot can interact with its environment using a model of the environment (map information) which is provided by map building mechanism. These two capabilities depends on each other and simultaneous operation of them is called SLAM (Simultaneous Localization and Map Building). While various sensors are used for this algorithm, vision-based approaches are relatively new and have attracted more interest in recent years. In this thesis work, a versatile Visual SLAM system is constructed and presented. In the core of this work is a vision-based simultaneous localization and map building algorithm which uses point features in the environment as visual landmarks and Extended Kalman Filter for state estimation. A detailed analysis of this algorithm is made including state estimation, feature extraction and data association steps. The algorithm is extended to be used for both stereo and single camera systems. The core of both algorithms is same and we mention the differences of both algorithms originated from the measurement dissimilarity. The algorithm is run also in different motion modes, namely predefined, manual and autonomous. Secondly, a map management strategy is developed especially for extended environments. When the robot runs the SLAM algorithm in large environments, the constructed map contains a great number of landmarks obviously. The efficiency algorithm takes part, when the total number of features exceeds a critical value for the system. In this case, the current map is rarefied without losing the geometrical distribution of the landmarks. Furthermore, a well-organized graphical user interface is implemented which enables the operator to select operational modes, change various parameters of the main SLAM algorithm and see the results of the SLAM operation both textually and graphically. Finally, a basic mission concept is defined in our system, in order to illustrate what robot can do using the outputs of the SLAM algorithm. All of these ideas mentioned are implemented in this thesis, experiments are conducted using a real robot and the analysis results are discussed by comparing the algorithm outputs with ground-truth measurements.

<p>One of Saab Space’s products is the S19 guidance system for sounding rockets.Today this system is based on an inertial navigation system that blindly calculatesthe position of the rocket by integrating sensor readings with unknown bias. Thepurpose of this thesis is to integrate a Global Positioning System (GPS) receiverinto the guidance system to increase precision and robustness. There are mainlytwo problems involved in this integration. One is to integrate the GPS with sensorfusion into the existing guidance system. The seconds is to get the GPS satellitetracking to work under extremely high dynamics. The first of the two problems issolved by using an Extended Kalman filter (EKF) with two different linearizations.One of them is uses Euler angles and the other is done with quaternions. Theintegration technique implemented in this thesis is a loose integration between theGPS receiver and the inertial navigation system. The main task of the EKF isto estimate the bias of the inertial navigation system sensors and correct it toeliminate drift in the position. The solution is verified by computing the positionof a car using a GPS and an inertial measurement unit. Different solutions to theGPS tracking problem are proposed in a pre-study.</p> / <p>En av Saab Space produkter är navigationssystemet S19 som styr sondraketer.Fram till idag har systemet varit baserat på ett tröghetsnavigeringssystem somblint räknar ut position genom att integrera tröghetsnavigerinssystemets sensorermed okända biaser. Syftet med detta exjobb är att integrera en GPS med tröghetsnavigeringsystemetför att öka robusthet och precision. Det kan i huvudsak delasupp i två problem; att integrera en GPS-mottagare med det befintliga navigationsystemetmed användning utav sensorfusion, och att få satellitföljningen attfungera under extremt höga dynamiska förhållanden. Det första av de två problemenlöses genom ett Extended Kalman filter (EKF) med två olika linjäriseringar.Den första linjäriseringen är med Eulervinklar och är välbeprövad. Den andra ärmed kvaternioner. Integrationstekniken som implementeras i detta Examensarbeteär en lös integration mellan GPS-mottagaren och tröghetsnavigeringssystemet. Huvudsyftetmed EKF:en är att estimera bias i tröghetsnavigeringsystemets sensoreroch korrigera dem för att eliminera drifter i position. Lösningen verifieras genomatt räkna ut positionen för en bil med GPS och en inertiell mätenhet. Olika lösningartill satellitföljningen föreslås i en förstudie.</p>

A novel structural health assessment (SHA) technique is proposed. It is a finite element-based time domain nonlinear system identification technique. The procedure is developed in two stages to incorporate several desirable features and increase its implementation potential. First, a weighted global iteration with an objective function is introduced in the unscented Kalman filter (UKF) procedure in order to obtain stable, convergent, and optimal solution. Furthermore, it also improves the capability of the UKF procedure to identify a large structural system using only a short duration of responses measured at a limited number of dynamic degrees of freedom (DDOFs). The combined procedure is denoted as unscented Kalman filter with weighted global iteration (UKF-WGI). Then, UKF-WGI is integrated with iterative least-squares with unknown input (ILS-UI) in order to increase its implementation potential. The substructure concept is also incorporated in the procedure. The integrated procedure is denoted as unscented Kalman filter with unknown input and weighted global iteration (UKF-UI-WGI). The two most important features of the method are that it does not need information on input excitation and uses only limited number of noise-contaminated response information to identify structural systems. Also, the method is able to identify the defects at the local element level by tracking the changes in the stiffness of the structural elements in the finite element representation. The UKF-UI-WGI procedure is implemented in two stages. In Stage 1, based on the location of input excitation, the substructure is selected. Using only responses at all DDOFs in the substructure, ILS-UI can identify the input excitation time-histories, stiffness parameters of all the elements in the substructure, and two Rayleigh damping coefficients. The outcomes of the first stage are necessary to initiate UKF-WGI. Using the information from Stage 1, the stiffness parameters of all the elements in the structure are identified using UKF-WGI in Stage 2. To demonstrate the effectiveness of the procedure, health assessment of relatively large structural systems is presented. Small and relatively large defects are introduced at different locations in the structure and the capability of the method to detect the health of the structure is examined. The optimum number and location of measured responses are also investigated. It is demonstrated that the method is capable of identifying defect-free and defective states of the structures using minimum information. Furthermore, it can locate defect spot within a defective element accurately. The comparative studies are also conducted between the proposed methods and available methods in the literature. First, it is between the UKF-WGI and extended Kalman filter with weighted global iteration (EKF-WGI) procedure. Then, it is between UKF-UI-WGI and generalized iterative least-squares extended Kalman filter with unknown input (GILS-EKF-UI) procedure, developed earlier by the research team. It is demonstrated that the proposed UKF-based procedures are superior to the EKF-based procedures for SHA.

Autonomous features in vehicles is already a big part of the automobile area and now many companies are looking for ways to make vehicles fully autonomous. Autonomous vehicles need to get information about the surrounding environment. The information is extracted from exteroceptive sensors and today vehicles often use laser scanners for this purpose. Laser scanners are very expensive and fragile, it is therefore interesting to investigate if cheaper radar sensors could be used. One big challenge when it comes to autonomous vehicles is to be able to use the exteroceptive sensors and extract a position of the vehicle and at the same time get a map of the environment. The area of Simultaneous Localization and Mapping (SLAM) is a well explored area when using laser scanners but is not that well explored when using radars. It has been investigated if it is possible to use radar sensors on a truck to create a map of the area where the truck drives. The truck has been equipped with ego-motion sensors and radars and the data from them has been fused together to get a position of the truck and to get a map of the surrounding environment, i.e. a SLAM algorithm has been implemented. The map is represented by an Occupancy Grid Map (OGM) which should only consist of static objects. The OGM is updated probabilistically by using a binary Bayes filter. To localize the truck with help of motion sensors an Extended Kalman Filter (EKF) is used together with a map and a scan match method. All these methods are put together to create a SLAM algorithm. A range rate filter method is used to filter out noise and non-static measurements from the radar. The results of this thesis show that it is possible to use radar sensors to create a map of a truck's surroundings. The quality of the map is considered to be good and details such as space between parked trucks, signs and light posts can be distinguished. It has also been proven that methods with low performance on their own can together with other methods work very well in the SLAM algorithm. Overall the SLAM algorithm works well but when driving in unexplored areas with a low number of objects problems with positioning might occur. A real time system has also been implemented and the map can be seen at the same time as the truck is manoeuvred.

O presente trabalho trata da criação de uma nova abordagem para predição de séries temporais ruidosas, com modelo desconhecido e que apresentam grandes não-linearidades. O novo método neuro-estatístico proposto combina uma rede neural de múltiplas camadas com o método estatístico Filtro de Kalman Estendido. A justificativa para a junção dessas abordagens é o fato de possuírem características complementares para o tratamento das peculiaridades das séries descritas. Quanto ao ruído, o FKE consegue minimizar a sua influência, trabalhando com a variância do ruído extraído dos dados reais. Quanto ao modelo gerador da série, as redes neurais aproximam a sua função, aprendendo a partir de amostras dos próprios dados. Grandes não-linearidades também são tratadas pelas RNs. O método neuro-estatístico segue a estrutura do FKE, utilizando a RN como processo preditivo. Com isso, elimina-se a necessidade de conhecimento prévio da função de transição de estados. O poder de tratamento de não-linearidades da RN é mantido, utilizando-se a previsão desta como estimativa de estado e os seus valores internos para cálculo das jacobianas do FKE. As matrizes de covariâncias dos erros de estimativa e dos ruídos são utilizadas para melhora do resultado obtido pela RN. A rede é treinada com um conjunto de dados retirado do histórico da série, de maneira off-line, possibilitando o uso de poderosas estruturas de redes de múltiplas camadas. Os resultados do método neuro-estatístico são comparados com a mesma configuração de RN utilizada em sua composição, sendo ambos aplicados na série caótica de Mackey-Glass e em uma série combinada de senos. Ambas séries possuem grandes não-linearidades e são acrescidas de ruído. O novo método alcança resultados satisfatórios, melhorando o resultado da RN em todos os experimentos. Também são dadas contribuições no ajuste dos parâmetros do FKE, utilizados no novo método. O método híbrido proporciona uma melhora mútua entre a RN e o FKE, explicando os bons resultados obtidos. / This work presents a new forecast method over highly nonlinear noisy time series. The neural statistical method uses a multi-layer perceptron (NN) and the Extended Kalman Filter (EKF). The justification for the combination of these approaches is that they possess complementary characteristics for the treatment of the peculiarities of the series. The EKF minimizes the influence of noise, working with the variance of the noise obtained from the real data. The NN approximates the generating model’s function. High nonlinearities are also treated by the neural network. The neural statistical method follows the structure of the EKF, using the NN as the predictive process. Thus, it isn’t necessary previous knowledge of the state transition function. The power of treatment of nonlinearities of the NN is kept, using forecast of this as estimative of state and its internal values for calculation of the Jacobian matrix of the EKF. The error estimative covariance and the noise covariance matrixes are used to improve the NN outcome. The NN is trained offline by past observations of the series, which enable the use of powerfuls neural networks. The results of the neural statistical method are compared with the same configuration of NN used in its composition, being applied in the chaotic series of Mackey-Glass and an sine mistures series. Both series are noisy and highly nonlinear. The new method obtained satisfactory result, improving the result of the regular NN in all experiments. The method also contributes in the adjustment of the parameters of the EKF. The hybrid method has a mutual improvement between the NN and the EKF, which explains the obtained good results.

A retroalimentação de controladores preditivos que utilizam modelos em espaço de estado pode ser realizada de duas formas: (a) correção por bias, em que as saídas preditas são corrigidas adicionando-se um valor proporcional a discrepância encontrada entre o valor medido atual e sua respectiva predição e por (b) retroalimentação dos estados, onde se determinam as condições iniciais através da estimação dos estados, e a partir de uma melhor condição inicial se realizam as predições futuras usadas no cálculo das ações de controle. Nesta dissertação estas duas abordagens são comparadas utilizando a Planta Laboratorial de Seis Tanques Esféricos. As técnicas de Filtro de Kalman Estendido (EKF) e Filtro de Kalman Estendido com Restrições (CEKF) foram empregadas para estimar os estados não medidos. Inicialmente foram feitos testes off-line destes algoritmos de estimação. Para estes testes são utilizados uma série de dados da planta laboratorial do estudo de caso, na qual são estudadas as influências de diversos fatores de ajuste que determinam a qualidade final de estimação. Estes ajustes serviram de base para a aplicação destes algoritmos em tempo real, quando então, estimadores de estados estão associados ao sistema de controle do processo baseado em um algoritmo de controle preditivo. Após se ter certificado a qualidade das estimações de estado, partiu-se para sua utilização como uma alternativa de retroalimentação de controladores preditivos. Estes resultados foram comparados com os obtidos através da correção simples por bias. Os resultados experimentais apontam para uma marginal piora devido à retroalimentação por estimadores de estados frente à correção por bias, pelo menos para o caso do controlador preditivo linear utilizado na comparação. Entretanto, espera-se que resultados melhores sejam obtidos no caso de modelos preditivos não-lineares, uma vez que nestes casos o modelo é bem mais sensível à qualidade da condição inicial. / The feedback of controllers that use predictive models in state space can be accomplished in two ways: (a) bias correction, where the predicted outputs are corrected by adding a value proportional to the discrepancy found between the current measurement and its respective prediction; and by (b) state feedback, which establishes the initial conditions through the states estimation, and from a better initial condition are carried out the future predictions used in the calculation of control. In this thesis these two approaches are compared using a Laboratorial Plant of Six Spherical Tanks. The techniques of Extended Kalman Filter (EKF) and Constraint Extended Kalman Filter (CEKF) were used to estimate the unmeasured states. Initially, tests were carried out off-line for theses estimation algorithms. For such testing are used a dataset of the plant in case study, in which are studied the influences of several adjustment factors that they determine the final quality of estimation. These adjustments were used of base for the application of these algorithms in real time, when then state estimators are associated with the system of process control based on a predictive control algorithm. After having ascertained the quality of the state estimates, begins its use as an alternative for feedback of predictive controllers. These results were compared with those obtained by the simple correction of bias. The experimental results show a marginal worsening due to feedback from state estimated compared with bias correction, at least for the case of linear predictive controller used in the comparison. However, one expects that better results will be obtained in the case of non-linear predictive models, since in these cases the model is much more sensitive to the quality of the initial condition.

Nesta dissertação é apresentada a comparação do desempenho de três estimadores - o Filtro de Kalman Estendido, o Filtro de Kalman Unscented e o Filtro de Partículas - aplicados para estimar a postura de um robô diferencial. Uma câmera foi fixa no teto para cobrir todo o campo operacional do robô durante os experimentos, a fim de extrair o mapa e gerar o ground truth. Isso permitiu realizar uma análise do erro de forma precisa a cada instante de tempo. O desempenho de cada um dos estimadores foi avaliado sistematicamente e numericamente para duas trajetórias. Os resultados desse primeiro experimento demonstram que os filtros proporcionam grandes melhorias em relação à odometria e que o modelo dos sensores é crítico para obter esse desempenho. O Filtro de Partículas mostrou um desempenho melhor em relação aos demais nos dois percursos. No entanto, seu elevado custo computacional dificulta sua implementação em uma aplicação de tempo real. O Filtro de Kalman Unscented, por sua vez, mostrou um desempenho semelhante ao Filtro de Kalman Estendido durante a primeira trajetória. Porém, na segunda trajetória, a qual possui uma quantidade maior de curvas, o Filtro de Kalman Unscented mostrou uma melhora significativa em relação ao Filtro de Kalman Estendido. Foi realizado um segundo experimento, em que o robô planeja e executa duas trajetórias. Os resultados obtidos mostraram que o robô consegue chegar a um determinado local com uma precisão da mesma ordem de grandeza do que a obtida durante a estimação de estados do robô. / In this dissertation, the performance of three nonlinear-model based estimators - the Extended Kalman Filter, the Unscented Kalman Filter and the Particle Filter - applied to pose estimation of a differential drive robot is compared. A camera was placed above the operating field of the robot to record the experiments in order to extract the map and generate the ground truth so the evaluation of the error can be done at each time step with high accuracy. The performance of each estimator is assessed systematically and numerically for two robot trajectories. The first experimental results showed that all estimators provide large improvements with respect to odometry and that the sensor modeling is critical for their performance. The particle filter showed a better performance than the others on both experiments, however, its high computational cost makes it difficult to implement in a real-time application. The Unscented Kalman Filter showed a similar performance to the Extended Kalman Filter during the first trajectory. However, during the second one (a curvier path) the Unscented Kalman Filter showed a significant improvement over the Extended Kalman Filter. A second experiment was carried out where the robot plans and executes a trajectory. The results showed the robot can reach a predefined location with an accuracy of the same order of magnitude as the obtained during the robot pose estimation.

O presente trabalho trata da criação de uma nova abordagem para predição de séries temporais ruidosas, com modelo desconhecido e que apresentam grandes não-linearidades. O novo método neuro-estatístico proposto combina uma rede neural de múltiplas camadas com o método estatístico Filtro de Kalman Estendido. A justificativa para a junção dessas abordagens é o fato de possuírem características complementares para o tratamento das peculiaridades das séries descritas. Quanto ao ruído, o FKE consegue minimizar a sua influência, trabalhando com a variância do ruído extraído dos dados reais. Quanto ao modelo gerador da série, as redes neurais aproximam a sua função, aprendendo a partir de amostras dos próprios dados. Grandes não-linearidades também são tratadas pelas RNs. O método neuro-estatístico segue a estrutura do FKE, utilizando a RN como processo preditivo. Com isso, elimina-se a necessidade de conhecimento prévio da função de transição de estados. O poder de tratamento de não-linearidades da RN é mantido, utilizando-se a previsão desta como estimativa de estado e os seus valores internos para cálculo das jacobianas do FKE. As matrizes de covariâncias dos erros de estimativa e dos ruídos são utilizadas para melhora do resultado obtido pela RN. A rede é treinada com um conjunto de dados retirado do histórico da série, de maneira off-line, possibilitando o uso de poderosas estruturas de redes de múltiplas camadas. Os resultados do método neuro-estatístico são comparados com a mesma configuração de RN utilizada em sua composição, sendo ambos aplicados na série caótica de Mackey-Glass e em uma série combinada de senos. Ambas séries possuem grandes não-linearidades e são acrescidas de ruído. O novo método alcança resultados satisfatórios, melhorando o resultado da RN em todos os experimentos. Também são dadas contribuições no ajuste dos parâmetros do FKE, utilizados no novo método. O método híbrido proporciona uma melhora mútua entre a RN e o FKE, explicando os bons resultados obtidos. / This work presents a new forecast method over highly nonlinear noisy time series. The neural statistical method uses a multi-layer perceptron (NN) and the Extended Kalman Filter (EKF). The justification for the combination of these approaches is that they possess complementary characteristics for the treatment of the peculiarities of the series. The EKF minimizes the influence of noise, working with the variance of the noise obtained from the real data. The NN approximates the generating model’s function. High nonlinearities are also treated by the neural network. The neural statistical method follows the structure of the EKF, using the NN as the predictive process. Thus, it isn’t necessary previous knowledge of the state transition function. The power of treatment of nonlinearities of the NN is kept, using forecast of this as estimative of state and its internal values for calculation of the Jacobian matrix of the EKF. The error estimative covariance and the noise covariance matrixes are used to improve the NN outcome. The NN is trained offline by past observations of the series, which enable the use of powerfuls neural networks. The results of the neural statistical method are compared with the same configuration of NN used in its composition, being applied in the chaotic series of Mackey-Glass and an sine mistures series. Both series are noisy and highly nonlinear. The new method obtained satisfactory result, improving the result of the regular NN in all experiments. The method also contributes in the adjustment of the parameters of the EKF. The hybrid method has a mutual improvement between the NN and the EKF, which explains the obtained good results.

A retroalimentação de controladores preditivos que utilizam modelos em espaço de estado pode ser realizada de duas formas: (a) correção por bias, em que as saídas preditas são corrigidas adicionando-se um valor proporcional a discrepância encontrada entre o valor medido atual e sua respectiva predição e por (b) retroalimentação dos estados, onde se determinam as condições iniciais através da estimação dos estados, e a partir de uma melhor condição inicial se realizam as predições futuras usadas no cálculo das ações de controle. Nesta dissertação estas duas abordagens são comparadas utilizando a Planta Laboratorial de Seis Tanques Esféricos. As técnicas de Filtro de Kalman Estendido (EKF) e Filtro de Kalman Estendido com Restrições (CEKF) foram empregadas para estimar os estados não medidos. Inicialmente foram feitos testes off-line destes algoritmos de estimação. Para estes testes são utilizados uma série de dados da planta laboratorial do estudo de caso, na qual são estudadas as influências de diversos fatores de ajuste que determinam a qualidade final de estimação. Estes ajustes serviram de base para a aplicação destes algoritmos em tempo real, quando então, estimadores de estados estão associados ao sistema de controle do processo baseado em um algoritmo de controle preditivo. Após se ter certificado a qualidade das estimações de estado, partiu-se para sua utilização como uma alternativa de retroalimentação de controladores preditivos. Estes resultados foram comparados com os obtidos através da correção simples por bias. Os resultados experimentais apontam para uma marginal piora devido à retroalimentação por estimadores de estados frente à correção por bias, pelo menos para o caso do controlador preditivo linear utilizado na comparação. Entretanto, espera-se que resultados melhores sejam obtidos no caso de modelos preditivos não-lineares, uma vez que nestes casos o modelo é bem mais sensível à qualidade da condição inicial. / The feedback of controllers that use predictive models in state space can be accomplished in two ways: (a) bias correction, where the predicted outputs are corrected by adding a value proportional to the discrepancy found between the current measurement and its respective prediction; and by (b) state feedback, which establishes the initial conditions through the states estimation, and from a better initial condition are carried out the future predictions used in the calculation of control. In this thesis these two approaches are compared using a Laboratorial Plant of Six Spherical Tanks. The techniques of Extended Kalman Filter (EKF) and Constraint Extended Kalman Filter (CEKF) were used to estimate the unmeasured states. Initially, tests were carried out off-line for theses estimation algorithms. For such testing are used a dataset of the plant in case study, in which are studied the influences of several adjustment factors that they determine the final quality of estimation. These adjustments were used of base for the application of these algorithms in real time, when then state estimators are associated with the system of process control based on a predictive control algorithm. After having ascertained the quality of the state estimates, begins its use as an alternative for feedback of predictive controllers. These results were compared with those obtained by the simple correction of bias. The experimental results show a marginal worsening due to feedback from state estimated compared with bias correction, at least for the case of linear predictive controller used in the comparison. However, one expects that better results will be obtained in the case of non-linear predictive models, since in these cases the model is much more sensitive to the quality of the initial condition.

abstract: Toward the ambitious long-term goal of a fleet of cooperating Flexible Autonomous Machines operating in an uncertain Environment (FAME), this thesis addresses various perception and control problems in autonomous aerial robotics. The objective of this thesis is to motivate the use of perspective cues in single images for the planning and control of quadrotors in indoor environments. In addition to providing empirical evidence for the abundance of such cues in indoor environments, the usefulness of these perspective cues is demonstrated by designing a control algorithm for navigating a quadrotor in indoor corridors. An Extended Kalman Filter (EKF), implemented on top of the vision algorithm, serves to improve the robustness of the algorithm to changing illumination. In this thesis, vanishing points are the perspective cues used to control and navigate a quadrotor in an indoor corridor. Indoor corridors are an abundant source of parallel lines. As a consequence of perspective projection, parallel lines in the real world, that are not parallel to the plane of the camera, intersect at a point in the image. This point is called the vanishing point of the image. The vanishing point is sensitive to the lateral motion of the camera and hence the quadrotor. By tracking the position of the vanishing point in every image frame, the quadrotor can navigate along the center of the corridor. Experiments are conducted using the Augmented Reality (AR) Drone 2.0. The drone is equipped with the following componenets: (1) 720p forward facing camera for vanishing point detection, (2) 240p downward facing camera, (3) Inertial Measurement Unit (IMU) for attitude control , (4) Ultrasonic sensor for estimating altitude, (5) On-board 1 GHz Processor for processing low level commands. The reliability of the vision algorithm is presented by flying the drone in indoor corridors. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2018