Simultaneous Localization and Mapping (SLAM) is a recursive probabilistic inferencing process used for robot navigation when Global Positioning Systems (GPS) are unavailable. SLAM operates by building a map of the robot environment, while concurrently localizing the robot within this map. The ultimate goal of SLAM is to operate anywhere using the environment's natural features as landmarks. Such a goal is difficult to achieve for several reasons. Firstly, different environments contain different types of natural features, each exhibiting large variance in its shape and appearance. Secondly, objects look differently from different viewpoints and it is therefore difficult to always recognize them. Thirdly, in most outdoor environments it is not possible to predict the motion of a vehicle using wheel encoders because of errors caused by slippage. Finally, the design of a SLAM system to operate in a large-scale outdoor setting is in itself a challenge. <br /><br /> The above issues are addressed as follows. Firstly, a camera is used to recognize the environmental context (e. g. , indoor office, outdoor park) by analyzing the holistic spectral content of images of the robot's surroundings. A type of feature (e. g. , trees for a park) is then chosen for SLAM that is likely observable in the recognized setting. A novel tree detection system is introduced, which is based on perceptually organizing the content of images into quasi-vertical structures and marking those structures that intersect ground level as tree trunks. Secondly, a new tree recognition system is proposed, which is based on extracting Scale Invariant Feature Transform (SIFT) features on each tree trunk region and matching trees in feature space. Thirdly, dead-reckoning is performed via an Inertial Navigation System (INS), bounded by non-holonomic constraints. INS are insensitive to slippage and varying ground conditions. Finally, the developed Computer Vision and Inertial systems are integrated within the framework of an Extended Kalman Filter into a working Vision-INS SLAM system, named VisSLAM. <br /><br /> VisSLAM is tested on data collected during a real test run in an outdoor unstructured environment. Three test scenarios are proposed, ranging from semi-automatic detection, recognition, and initialization to a fully automated SLAM system. The first two scenarios are used to verify the presented inertial and Computer Vision algorithms in the context of localization, where results indicate accurate vehicle pose estimation for the majority of its journey. The final scenario evaluates the application of the proposed systems for SLAM, where results indicate successful operation for a long portion of the vehicle journey. Although the scope of this thesis is to operate in an outdoor park setting using tree trunks as landmarks, the developed techniques lend themselves to other environments using different natural objects as landmarks.

Simultaneous Localization and Mapping (SLAM) is a recursive probabilistic inferencing process used for robot navigation when Global Positioning Systems (GPS) are unavailable. SLAM operates by building a map of the robot environment, while concurrently localizing the robot within this map. The ultimate goal of SLAM is to operate anywhere using the environment's natural features as landmarks. Such a goal is difficult to achieve for several reasons. Firstly, different environments contain different types of natural features, each exhibiting large variance in its shape and appearance. Secondly, objects look differently from different viewpoints and it is therefore difficult to always recognize them. Thirdly, in most outdoor environments it is not possible to predict the motion of a vehicle using wheel encoders because of errors caused by slippage. Finally, the design of a SLAM system to operate in a large-scale outdoor setting is in itself a challenge. <br /><br /> The above issues are addressed as follows. Firstly, a camera is used to recognize the environmental context (e. g. , indoor office, outdoor park) by analyzing the holistic spectral content of images of the robot's surroundings. A type of feature (e. g. , trees for a park) is then chosen for SLAM that is likely observable in the recognized setting. A novel tree detection system is introduced, which is based on perceptually organizing the content of images into quasi-vertical structures and marking those structures that intersect ground level as tree trunks. Secondly, a new tree recognition system is proposed, which is based on extracting Scale Invariant Feature Transform (SIFT) features on each tree trunk region and matching trees in feature space. Thirdly, dead-reckoning is performed via an Inertial Navigation System (INS), bounded by non-holonomic constraints. INS are insensitive to slippage and varying ground conditions. Finally, the developed Computer Vision and Inertial systems are integrated within the framework of an Extended Kalman Filter into a working Vision-INS SLAM system, named VisSLAM. <br /><br /> VisSLAM is tested on data collected during a real test run in an outdoor unstructured environment. Three test scenarios are proposed, ranging from semi-automatic detection, recognition, and initialization to a fully automated SLAM system. The first two scenarios are used to verify the presented inertial and Computer Vision algorithms in the context of localization, where results indicate accurate vehicle pose estimation for the majority of its journey. The final scenario evaluates the application of the proposed systems for SLAM, where results indicate successful operation for a long portion of the vehicle journey. Although the scope of this thesis is to operate in an outdoor park setting using tree trunks as landmarks, the developed techniques lend themselves to other environments using different natural objects as landmarks.

Doctor of Philosophy (PhD) / This thesis addresses the problem of building statistical models for multi-sensor perception in unstructured outdoor environments. The perception problem is divided into three distinct tasks: recognition, representation and association. Recognition is cast as a statistical classification problem where inputs are images or a combination of images and ranging information. Given the complexity and variability of natural environments, this thesis investigates the use of Bayesian statistics and supervised dimensionality reduction to incorporate prior information and fuse sensory data. A compact probabilistic representation of natural objects is essential for many problems in field robotics. This thesis presents techniques for combining non-linear dimensionality reduction with parametric learning through Expectation Maximisation to build general representations of natural features. Once created these models need to be rapidly processed to account for incoming information. To this end, techniques for efficient probabilistic inference are proposed. The robustness of localisation and mapping algorithms is directly related to reliable data association. Conventional algorithms employ only geometric information which can become inconsistent for large trajectories. A new data association algorithm incorporating visual and geometric information is proposed to improve the reliability of this task. The method uses a compact probabilistic representation of objects to fuse visual and geometric information for the association decision. The main contributions of this thesis are: 1) a stochastic representation of objects through non-linear dimensionality reduction; 2) a landmark recognition system using a visual and ranging sensors; 3) a data association algorithm combining appearance and position properties; 4) a real-time algorithm for detection and segmentation of natural objects from few training images and 5) a real-time place recognition system combining dimensionality reduction and Bayesian learning. The theoretical contributions of this thesis are demonstrated with a series of experiments in unstructured environments. In particular, the combination of recognition, representation and association algorithms is applied to the Simultaneous Localisation and Mapping problem (SLAM) to close large loops in outdoor trajectories, proving the benefits of the proposed methodology.

This thesis consists of two parts united by the theme of path/action planning for a mobile robot. Part I presents the Second Opinion Planner (SOP), and Part II presents a new paradigm for navigating, growing, and planning on a Network of Reusable Paths (NRP). Path/action planning is common to both parts in that the planning algorithm must choose the terrain assessment or localization technique at the path-planning stage. Terrain-assessment algorithms follow the trend of low-fidelity at low-cost and high-fidelity at high-cost. Using a high-fidelity method on all the raw terrain data can drastically increase a robot's total path cost (cost of driving, planning, and doing the terrain assessment). SOP is a path-planning algorithm that uses a hierarchy of terrain-assessment methods, from low-fidelity to high-fidelity, and seeks to limit high-cost assessment to areas where it is beneficial. The decision to assess some terrain with a higher-fidelity method is made considering potential path benefits and the cost of assessment. SOP provides a means to triage large amounts of terrain data. The system is demonstrated on simulated problems and in real terrain from an experimental field test carried out on Devon Island, Canada. The SOP plans are quite close to the minimum possible cost. Growing a NRP is an approach to navigation that allows a mobile robot to autonomously explore unmapped, GPS-denied environments. This new paradigm results in closer goal acquisition and a more robust approach to exploration with a mobile robot, when compared to a classic approach to guidance, navigation, and control. A NRP is a simple Simultaneous Localization And Mapping system that can be shown to be a physical embodiment of a Rapidly-exploring Random Tree planner. Simulation results are presented, as well as the results from two different robotic test systems that were tested in planetary analogue environments. NRP offers benefits to planetary exploration by allowing a rover to be used for the parallel scientific investigations. This increases the number of sites that can be investigated in a short time, as compared to a serial approach to exploration. Two mock missions were carried out at planetary analogue sites.

This thesis consists of two parts united by the theme of path/action planning for a mobile robot. Part I presents the Second Opinion Planner (SOP), and Part II presents a new paradigm for navigating, growing, and planning on a Network of Reusable Paths (NRP). Path/action planning is common to both parts in that the planning algorithm must choose the terrain assessment or localization technique at the path-planning stage. Terrain-assessment algorithms follow the trend of low-fidelity at low-cost and high-fidelity at high-cost. Using a high-fidelity method on all the raw terrain data can drastically increase a robot's total path cost (cost of driving, planning, and doing the terrain assessment). SOP is a path-planning algorithm that uses a hierarchy of terrain-assessment methods, from low-fidelity to high-fidelity, and seeks to limit high-cost assessment to areas where it is beneficial. The decision to assess some terrain with a higher-fidelity method is made considering potential path benefits and the cost of assessment. SOP provides a means to triage large amounts of terrain data. The system is demonstrated on simulated problems and in real terrain from an experimental field test carried out on Devon Island, Canada. The SOP plans are quite close to the minimum possible cost. Growing a NRP is an approach to navigation that allows a mobile robot to autonomously explore unmapped, GPS-denied environments. This new paradigm results in closer goal acquisition and a more robust approach to exploration with a mobile robot, when compared to a classic approach to guidance, navigation, and control. A NRP is a simple Simultaneous Localization And Mapping system that can be shown to be a physical embodiment of a Rapidly-exploring Random Tree planner. Simulation results are presented, as well as the results from two different robotic test systems that were tested in planetary analogue environments. NRP offers benefits to planetary exploration by allowing a rover to be used for the parallel scientific investigations. This increases the number of sites that can be investigated in a short time, as compared to a serial approach to exploration. Two mock missions were carried out at planetary analogue sites.

Doctor of Philosophy (PhD) / This thesis addresses the problem of building statistical models for multi-sensor perception in unstructured outdoor environments. The perception problem is divided into three distinct tasks: recognition, representation and association. Recognition is cast as a statistical classification problem where inputs are images or a combination of images and ranging information. Given the complexity and variability of natural environments, this thesis investigates the use of Bayesian statistics and supervised dimensionality reduction to incorporate prior information and fuse sensory data. A compact probabilistic representation of natural objects is essential for many problems in field robotics. This thesis presents techniques for combining non-linear dimensionality reduction with parametric learning through Expectation Maximisation to build general representations of natural features. Once created these models need to be rapidly processed to account for incoming information. To this end, techniques for efficient probabilistic inference are proposed. The robustness of localisation and mapping algorithms is directly related to reliable data association. Conventional algorithms employ only geometric information which can become inconsistent for large trajectories. A new data association algorithm incorporating visual and geometric information is proposed to improve the reliability of this task. The method uses a compact probabilistic representation of objects to fuse visual and geometric information for the association decision. The main contributions of this thesis are: 1) a stochastic representation of objects through non-linear dimensionality reduction; 2) a landmark recognition system using a visual and ranging sensors; 3) a data association algorithm combining appearance and position properties; 4) a real-time algorithm for detection and segmentation of natural objects from few training images and 5) a real-time place recognition system combining dimensionality reduction and Bayesian learning. The theoretical contributions of this thesis are demonstrated with a series of experiments in unstructured environments. In particular, the combination of recognition, representation and association algorithms is applied to the Simultaneous Localisation and Mapping problem (SLAM) to close large loops in outdoor trajectories, proving the benefits of the proposed methodology.

In this dissertation, we seek to improve collision avoidance for autonomous underwater vehicles (AUVs). More specifically, we consider the case of a small AUV using a forward-looking sonar system with a limited number of beams. We describe a high-fidelity sonar model and simulation environment that was developed to aid in the design of the sonar system. The simulator achieves real-time visualization through ray tracing and approximation, and can be used to assess sonar design choices, such as beam pattern and beam location, and to evaluate obstacle detection algorithms. We analyze the benefit of using a few beams instead of a single beam for a low-cost obstacle avoidance sonar for small AUVs. Single-beam systems are small and low-cost, while multi-beam sonar systems are more expensive and complex, often incorporating hundreds of beams. We want to quantify the improvement in obstacle avoidance performance of adding a few beams to a single-beam system. Furthermore, we developed a collision avoidance strategy specifically designed for the novel sonar system. The collision avoidance strategy is based on posterior expected loss, and explicitly couples obstacle detection, collision avoidance, and planning. We demonstrate the strategy with field trials using the 690 AUV, built by the Center for Marine Autonomy and Robotics at Virginia Tech, with a prototype forward-looking sonar comprising of nine beams. / Doctor of Philosophy / This dissertation focuses on improving collision avoidance capabilities for small autonomous underwater vehicles (AUVs). Specifically, we are looking at the scenario of an AUV equipped with a forward-looking sonar system using only a few beams to detect obstacles in our environment. We develop a sophisticated sonar model and simulation environment to facilitate the design of the sonar system. Our simulator enables real-time visualization, offering insights into sonar design aspects. It also serves as a tool for evaluating obstacle detection algorithms. The research investigates the advantages of utilizing multiple beams compared to a single-beam system for a cost-effective obstacle avoidance solution for small AUVs. Single-beam sonar systems are small and affordable, while multi-beam sonar systems are more complex and expensive. The aim is to quantify the improvement in obstacle avoidance performance when adding additional sonar beams. Additionally, a collision avoidance strategy tailored to the novel sonar system is developed. This strategy, developed using a statistical model, integrates obstacle detection, collision avoidance, and planning. The effectiveness of the strategy is demonstrated through field trials using the 690 AUV, constructed by the Center for Marine Autonomy and Robotics at Virginia Tech, equipped with a prototype forward-looking sonar using nine beams.

Many mobile robot applications require robots to act safely and intelligently in complex unfamiliarenvironments with little structure and limited or unavailable human supervision. As arobot is forced to operate in an environment that it was not engineered or trained for, various aspectsof its performance will inevitably degrade. Roboticists equip robots with powerful sensorsand data sources to deal with uncertainty, only to discover that the robots are able to make onlyminimal use of this data and still find themselves in trouble. Similarly, roboticists develop andtrain their robots in representative areas, only to discover that they encounter new situations thatare not in their experience base. Small problems resulting in mildly sub-optimal performance areoften tolerable, but major failures resulting in vehicle loss or compromised human safety are not.This thesis presents a series of online algorithms to enable a mobile robot to better deal withuncertainty in unfamiliar domains in order to improve its navigational abilities, better utilizeavailable data and resources and reduce risk to the vehicle. We validate these algorithms throughextensive testing onboard large mobile robot systems and argue how such approaches can increasethe reliability and robustness of mobile robots, bringing them closer to the capabilitiesrequired for many real-world applications.

Achieving robust and reliable autonomous operation even in complex unstructured environments is a central goal of field robotics. As the environments and scenarios to which robots are applied have continued to grow in complexity, so has the challenge of properly defining preferences and tradeoffs between various actions and the terrains they result in traversing. These definitions and parameters encode the desired behavior of the robot; therefore their correctness is of the utmost importance. Current manual approaches to creating and adjusting these preference models and cost functions have proven to be incredibly tedious and time-consuming, while typically not producing optimal results except in the simplest of circumstances. This thesis presents the development and application of machine learning techniques that automate the construction and tuning of preference models within complex mobile robotic systems. Utilizing the framework of inverse optimal control, expert examples of robot behavior can be used to construct models that generalize demonstrated preferences and reproduce similar behavior. Novel learning from demonstration approaches are developed that offer the possibility of significantly reducing the amount of human interaction necessary to tune a system, while also improving its final performance. Techniques to account for the inevitability of noisy and imperfect demonstration are presented, along with additional methods for improving the efficiency of expert demonstration and feedback. The effectiveness of these approaches is confirmed through application to several real world domains, such as the interpretation of static and dynamic perceptual data in unstructured environments and the learning of human driving styles and maneuver preferences. Extensive testing and experimentation both in simulation and in the field with multiple mobile robotic systems provides empirical confirmation of superior autonomous performance, with less expert interaction and no hand tuning. These experiments validate the potential applicability of the developed algorithms to a large variety of future mobile robotic systems.

Arctic regions are thought to be more sensitive to climate change fluctuations, making weather data from these regions more valuable for climate modeling. Scientists have expressed an interest in deploying a robotic sensor network in these areas, minimizing the exposure of human researchers to the harsh environment, while allowing dense, targeted data collection to commence. For any such robotic system to be successful, a certain set of base navigational functionality must be developed. Further, these navigational algorithms must rely on the types of low-cost sensors that would be viable for use in a multi-agent system. A set of vision-based processing techniques have been proposed, which augment current robotic technologies for use in glacial terrains. Specifically, algorithms for estimating terrain traversability, robot localization, and terrain reconstruction have been developed which use data collected exclusively from a single camera and other low-cost robotic sensors. For traversability assessment, a custom algorithm was developed that uses local scale surface texture to estimate the terrain slope. Additionally, a horizon line estimation system has been proposed that is capable of coping with low-contrast, ambiguous horizons. For localization, a monocular simultaneous localization and mapping (SLAM) filter has been fused with consumer-grade GPS measurements to produce full robot pose estimates that do not drift over long traverses. Finally, a terrain reconstruction methodology has been proposed that uses a Gaussian process framework to incorporate sparse SLAM landmarks with dense slope estimates to produce a single, consistent terrain model. These algorithms have been tested within a custom glacial terrain computer simulation and against multiple data sets acquired during glacial field trials. The results of these tests indicate that vision is a viable sensing modality for autonomous glacial robotics, despite the obvious challenges presented by low-contrast glacial scenery. The findings of this work are discussed within the context of the larger arctic sensor network project, and a direction for future work is recommended.