Telemanipulators allow a user to interact with a potentially dangerous environment remotely. Deploying a robot arm from a UAV would allow an operator to reach farther and quicker than he or she would with a ground robotics system alone. This thesis will discuss the design and fabrication of a compact, light, 3 degree of freedom robot arm using common off the shelf products and machined components that in combination can pick up half pound samples and has a reach of 260 mm. Also addressed is making the telemanipulator interface easier to use. One of the challenges in using a robot arm with a single camera in a beyond line-of-sight scenario is the difficulty of interacting with the environment because of a loss of depth information. This lack of information can be remedied with additional sensors. Once depth to an object of interest is known, the sampler can automatically pick up objects of interest. The manipulator arm will be used in conjunction with systems developed by the Unmanned Systems Laboratory at Virginia Tech. This group is developing a unmanned ground vehicle to be carried in the payload pod of a unmanned aerial vehicle. The robot's ultimate objective is to collect shrapnel and bomb material from potentially dangerous environments. / Master of Science

This thesis explores three topics in the field of cost estimation for Unmanned Autonomous Systems. First, we propose a common definition of an Unmanned Autonomous System. We accomplish this through exhausting the literature in the areas cost estimation, autonomy in its current form, and how such advanced systems might be integrated into their environment. Second, we introduce a method to estimate the cost of Unmanned Autonomous Systems utilizing existing parametric cost estimation tools: SEER–HDR, COCOMO II, COSYSMO, and two cost estimating relationships–weight and performance. This discussion is guided by focusing on how current tools attempt to account for emergent systems. We also attempt to address challenges surrounding autonomy. To address these challenges from a cost perspective, this thesis recommends modifications to parameters within COCOMO II–via the use of object-oriented function points in lieu of current methods, and COSYSMO–via the introduction of two cost drivers namely, TVED and HRI-T. Third, we conduct analysis on four current Army Unmanned Autonomous Systems in an attempt to establish early trends within existing estimates. Finally, we explore areas of further research and discuss the implications of how pursing a more adequate cost model will lead to a better understanding of this ill-defined paradigm. *This material is based upon work supported by the Naval Postgraduate School Acquisition Research Program under Grant No. N00244-15-1-0008. The views expressed in written materials or publications, and/or made by speakers, moderators, and presenters, do not necessarily reflect the official policies of the Naval Postgraduate School nor does mention of trade names, commercial practices, or organizations imply endorsement by the U.S. Government.

The development of Automated Driving Systems (ADS) has been ongoing for decades in varying levels of sophistication. Levels of automation are defined by Society of American Engineers (SAE) as 0 through 5, with 0 being full human control and 5 being full automation control. Another way to describe levels of automation is through concepts of Functional Safety (FuSa) and Operational Safety (OpSa). These terms of FuSa and OpSa are important, because ADS testing relies on both. Current recommendations for ADS testing include both OpSa and FuSa requirements. However, an examination of ADS safety requirements (e.g., industry reports, post-crash analysis reports, etc.) reveals that ADS safety arguments, in practice, depend almost completely on well-trained human operators, referred to in the industry as in vehicle fallback test drivers (IFTD). To date, the industry has never fielded a truly SAE L4 ADS on public roads due to this persistent hurdle of needing a human operator for Operational Safety. There is a tendency in ADS testing to reference International Standards Organization (ISOs) for validated vehicles for vehicles that are still in development (i.e., unvalidated). To be clear, ISOs for ADS end products are not necessarily applicable to ADS in development. With this in mind, there is a clear gap in the industry for unvalidated ADS literature. Because of this gap, ADS testing for unvalidated vehicles often relies on safety requirements for validated vehicles. This issue remains a significant challenge for ADS testing. Recognizing this gap in on-road, in-development vehicle safety, there is a need for the ADS industry to develop a clear strategy for transitioning from an IFTD (Operational Safety) to an ADS (Functional Safety). Therefore, the purpose of this thesis is to present a framework for transitioning from Operational Safety to Functional Safety. The framework makes this possible through an inductive analysis of available definitions of onroad safety to arrive at a definition that leverages Functional and Operational Safety along a continuum. Ultimately, the framework aims to contribute to onroad safety testing for the ADS industry. / Master of Science / The development of Self-Driving Cars has been ongoing for decades in varying levels of sophistication. Levels of automation are defined by Society of American Engineers (SAE) as 0 through 5, with 0 being full human control and 5 being full automation control. Another way to describe levels of automation is through concepts of Robotic Control and Human Control. If a vehicle relies completely on Human Control, a human operator is responsible for all on-road safety. On the other hand, a fully autonomous would be considered fully in Robotic Control. These terms of Robotic Control and Human Control are important, because Self-Driving Car testing relies on both. Current recommendations for Self-Driving Car testing include both Robotic Control and Human Control requirements. However, an examination of Self-Driving Cars documentation (e.g., industry reports, post-crash analysis reports, etc.) reveals that Self-Driving Car safety arguments, in practice, depend almost completely on well-trained human operators. To date, the industry has never fielded a truly SAE L4 Self-Driving Car on public roads due to this persistent hurdle of needing a human operator for Human Control. There is a tendency in Self-Driving Car testing to reference standars for validated vehicles for vehicles that are still in development (i.e., unvalidated). To be clear, standards for Self-Driving Car end products are not necessarily applicable to Self-Driving Cars in development. With this in mind, there is a clear gap in the industry for unvalidated Self-Driving Car literature. Because of this gap, Self-Driving Car testing for unvalidated vehicles often relies on documentation for validated vehicles. This issue remains a significant challenge for Self-Driving Car testing. Recognizing this gap in on-road, in-development vehicle safety, there is a need for the Self-Driving industry to develop a clear strategy for transitioning from Human Control to Robot Control. Therefore, the purpose of this thesis is to present a framework for transitioning from Human to Robot Control. The framework makes this possible through an inductive analysis of available definitions of onroad safety to arrive at a definition that leverages all definitions of Safety along a continuum. Ultimately, the framework aims to contribute to onroad safety testing for the Self-Driving industry.

published_or_final_version / abstract / Earth Sciences / Doctoral / Doctor of Philosophy

published_or_final_version / Earth Sciences / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / Earth Sciences / Master / Master of Philosophy

Autonomous Vehicles (AVs) should result in tremendous benefits to safe human transportation. Recent reports indicate a global average of 3,287 road crash related fatalities a day with the blame, in most cases, assigned to the human driver. By replacing the main cause, AVs are predicted to significantly reduce road accidents -- some claiming up to a 90% reduction on US roads. However, achieving these numbers is not simple. AVs are expected to assume tasks that human drivers perform both consciously and unconsciously -- in some instances, with Machine Learning. AVs incur new levels of complexity that, if handled incorrectly, can result in failures that cause loss of human life and damage to the environment. Accidents involving SAE Level 2 vehicles have highlighted such failures and demonstrated that AVs have a long way to go. The path towards safe AVs includes system architectures that provide effective failure monitoring, detection and mitigation. These architectures must produce AVs that degrade gracefully and remain sufficiently operational in the presence of failures. We introduce Safe-AV, a fault tolerant safety architecture for AVs that is based on the commonly adopted E-Gas 3 Level Monitoring Concept, the Simplex Architecture and guided by a thorough hazard analysis in the form of Systems-Theoretic Process Analysis (STPA). We commenced the architecture design with a review of some modern AV accidents which helped identify the types of failures AVs can present and acted as a first step to our STPA. The hazard analysis was applied to an initial AV architecture (without safety mechanisms) consisting of components that should be present in a typical AV (based on the literature and our ideas). Our STPA identified the system level accidents, hazards and corresponding loss scenarios that led to well-founded safety requirements which, in turn, evolved the initial architecture into Safe-AV. / Thesis / Master of Applied Science (MASc)

<p dir="ltr">Autonomous vehicles require motion forecasting of their surrounding multiagents (pedestrians</p><p dir="ltr">and vehicles) to make optimal decisions for navigation. The existing methods focus on</p><p dir="ltr">techniques to utilize the positions and velocities of these agents and fail to capture semantic</p><p dir="ltr">information from the scene. Moreover, to mitigate the increase in computational complexity</p><p dir="ltr">associated with the number of agents in the scene, some works leverage Euclidean distance to</p><p dir="ltr">prune far-away agents. However, distance-based metric alone is insufficient to select relevant</p><p dir="ltr">agents and accurately perform their predictions. To resolve these issues, we propose the</p><p dir="ltr">Semantics-aware Interactive Multiagent Motion Forecasting (SIMMF) method to capture</p><p dir="ltr">semantics along with spatial information and optimally select relevant agents for motion</p><p dir="ltr">prediction. Specifically, we achieve this by implementing a semantic-aware selection of relevant</p><p dir="ltr">agents from the scene and passing them through an attention mechanism to extract</p><p dir="ltr">global encodings. These encodings along with agents’ local information, are passed through</p><p dir="ltr">an encoder to obtain time-dependent latent variables for a motion policy predicting the future</p><p dir="ltr">trajectories. Our results show that the proposed approach outperforms state-of-the-art</p><p dir="ltr">baselines and provides more accurate and scene-consistent predictions. </p>

The purpose of this study was to investigate novel methods on an unmanned sailing boat, which enables it to sail fully autonomously, navigate safely, and perform long-term missions. The author used robotic sailing boat prototypes for field experiments as his main research method. Two robotic sailing boats have been developed especially for this purpose. A compact software model of a sailing boat's behaviour allowed for further evaluation of routing and obstacle avoidance methods in a computer simulation. The results of real-world experiments and computer simulations are validated against each other. It has been demonstrated that autonomous boat sailing is possible by the effective combination of appropriate new and novel techniques that will allow autonomous sailing boats to create appropriate routes, to react properly on obstacles and to carry out sailing manoeuvres by controlling rudder and sails. Novel methods for weather routing, collision avoidance, and autonomous manoeuvre execution have been proposed and successfully demonstrated. The combination of these techniques in a layered hybrid subsumption architecture make robotic sailing boats a promising tool for many applications, especially in ocean observation.

For adaptive motion control of an autonomous vehicle, operating in a generally structured environment, position and velocity feedback are required to ascertain the vehicle location relative to a reference. Whilst the literature offers techniques for guiding vehicles along external references, autonomous vehicles should be able to navigate between despatch locations without the need to rely on external guidance systems. Considerations of the vehicle stability and manoeuvrability favour a vehicle design with four independently steered wheels. A new motion control methodology has been proposed which utilises the geometric relationship of the angular displacements and the rotations of the wheels to estimate the longitudinal and lateral motions of the vehicle. The motion controller consists of three building blocks: the motion control system comprising the position tracking and the motion command generation; the electronic system comprising a data acquisition system and proprietary power electronics; the mechanical system which includes an undercarriage enabling permanent contact of the wheels with the floor. The components have been designed not only to perform optimally in their specific functions but also to ensure full compatibility within the integrated system. For reliable deduction of the wheel rotations with a high degree of accuracy a dedicated data acquisition interface has been developed, which enables data to be captured in parallel from four optical encoders mounted directly on the wheel axles. Parallel sampling of the angular wheel position and parallel actuation of all steering motors improves the accuracy of the system state and gives a higher degree of certainty. Considering only circular motion of the vehicle, a method for calculating the steering angles and wheel speeds based on the complex notation is presented. By cumulating the displacement vectors of the vehicle motion and the location of the centre of rotation between consecutive samples of the controller, the path of the vehicle is estimated. The difference between the nominal and the deduced centre of rotation is determined to minimise deviations from the reference trajectory and to allow the controller to adapt to changes in the road/tyre interface characteristics. The individual mechanical and electronic components have been assembled and tested. Additionally, the performance of the electronic interface has been evaluated on a purpose built test-bed. For the experimental validation of the methodology, a simple method of mapping the centre of curvature with a pen mounted at the nominal centre of rotation has been proposed. Experiments have been conducted with both the steering angles fixed to their theoretical values for the nominal centre of rotation and with the proportional steering controller enabled. The results from the latter method have shown a significantly reduced deviation from the nominal centre of rotation. The data captured of the angular wheel positions and steering angle settings has been analysed off-line. Good agreement is obtained between the deduced and the actual centres of rotation for the measurements averaged over 1.5 seconds. A number of different centres of rotation have been investigated and the required steering angles to compensate for the deviation have been plotted to form a control surface for the motion controller. The deviation between the estimated and the actual centre of curvature was less than 1.6% and adequate results could be obtained with the proportional steering controller.