Browsing by Author "Maciejewski, Anthony, committee member"
Now showing 1 - 12 of 12
Results Per Page
Sort Options
Item Open Access Algorithms in numerical algebraic geometry and applications(Colorado State University. Libraries, 2015) Hanson, Eric M., author; Bates, Daniel J., advisor; Peterson, Chris, committee member; Cavalieri, Renzo, committee member; Maciejewski, Anthony, committee memberThe topics in this dissertation, while independent, are unified under the field of numerical algebraic geometry. With ties to some of the oldest areas in mathematics, numerical algebraic geometry is relatively young as a field of study in its own right. The field is concerned with the numerical approximation of the solution sets of systems of polynomial equations and the manipulation of these sets. Given a polynomial system ƒ : CN → Cn, the methods of numerical algebraic geometry produce numerical approximations of the isolated solutions of ƒ(z) = 0, as well as points on any positive-dimensional components of the solution set, V(ƒ). In a short time, the work done in numerical algebraic geometry has significantly pushed the boundary of what is computable. This dissertation aims to further this work by contributing new algorithms to the field and using cutting edge techniques of the field to expand the scope of problems that can be addressed using numerical methods. We begin with an introduction to numerical algebraic geometry and subsequent chapters address independent topics: perturbed homotopies, exceptional sets and fiber products, and a numerical approach to finding unit distance embeddings of finite simple graphs. One of the most recent advances in numerical algebraic geometry is regeneration, an equation-by-equation homotopy method that is often more efficient than other approaches. However, the basic form of regeneration will not necessarily find all isolated singular solutions of a polynomial system without the additional cost of using deflation. In the second chapter, we present an alternative to deflation in the form of perturbed homotopies for solving polynomial systems. In particular, we propose first solving a perturbed version of the polynomial system, followed by a parameter homotopy to remove the perturbation. The aim of this chapter is two-fold. First, such perturbed homotopies are sometimes more efficient than regular homotopies, though they can also be less efficient. Second, a useful consequence is that the application of this perturbation to regeneration will yield all isolated solutions, including all singular isolated solutions. The third chapter considers families of polynomial systems which depend on parameters. There is a typical dimension for the variety defined by a system in the family; however, this dimension may jump for parameters in algebraic subsets of the parameter space. Sommese and Wampler exploited fiber products to give a numerical method for identifying these special parameter values. In this chapter, we propose a refined numerical approach to fiber products, which uses recent advancements in numerical algebraic geometry, such as regeneration extension. We show that this method is sometimes more efficient then known techniques. This gain in efficiency is due to the fact that regeneration extension allows the construction of the fiber product to be restricted to specified irreducible components. This work is motivated by applications in Kinematics - the study of mechanisms. As such we use an algebraic model of a two link arm to illustrate the algorithms developed in this chapter. The topic of the last chapter is the identification of unit distance embeddings of finite simple graphs. Given a graph G(V,E), a unit distance embedding is a map ɸ from the vertex set V into a metric space M such that if {vi,vj} is an element of E then the distance between ɸ (vi) and ɸ (vj) in M is one. Given G, we cast the question of the existence of a unit distance embedding in Rn as the question of the existence of a real solution to a system of polynomial equations. As a consequence, we are able to develop theoretic algorithms for determining the existence of a unit distance embedding and for determining the smallest dimension of Rn for which a unit distance embedding of G exists (that is, we determine the minimal embedding dimension of G). We put these algorithms into practice using the methods of numerical algebraic geometry. In particular, we consider unit distance embeddings of the Heawood Graph. This is the smallest example of a point-line incidence graph of a finite projective plan. In 1972, Chvátal conjectured that point-line incidence graphs of finite projective planes do not have unit-distance embeddings into R². In other words, Chvátal conjectured that the minimal embedding dimension of any point-line incidence graph of a finite projective plane is at least 3. We disprove this conjecture, adding hundreds of counterexamples to the 11 known counterexamples found by Gerbracht.Item Open Access Design, modeling, and optimization of 3D printed compliant mechanisms with applications to miniature walking robots(Colorado State University. Libraries, 2018) DeMario, Anthony R., author; Zhao, Jianguo, advisor; Stansloski, Mitchell, committee member; Maciejewski, Anthony, committee memberMiniature robots have many applications ranging from military surveillance to search and rescue assistance in disaster areas. Traditionally, fabrication of these robots has been labor intensive, time-consuming, and expensive. This thesis proposes to leverage recent advances in 3D printing technology to fabricate centimeter-scale walking robots utilizing compliant elements printed directly into the walking mechanisms in replacement of traditional revolute joints or rigid links. The ability to design around the capabilities of 3D printers and novel material choices gives miniature robots the ability to have multiple functions in the same mechanism, reduces the overall number of parts that must be assembled to make a functional robot, and decrease the time and cost of prototyping. This thesis details three areas of study for compliant mechanisms with applications to walking robots. First, we utilize multi-material 3D printing to fabricate a miniature walking robot (49 x 38 x 25mm) that directly replaces the traditional revolute joints in the designed walking mechanism with a custom, soft joint. Some links are also printed with soft materials to enhance the robustness and durability of the robot. Along with design and testing of the robot, we develop two numerical models to simulate the effects of the soft elements on the mechanism trajectory. Second, we leverage the numerical models to optimize the design of the walking mechanism to produce a trajectory similar to that of the same mechanism using all revolute joints. Third, we redesign the original robot to utilize a conductive polylactic acid (PLA) material to 3D print linkages that allow for changing joints locations by softening the desired part through applied electricity. This variable joint mechanism can create multiple trajectories without changing the mechanical structure, therefore creating a multi-functional compliant mechianism. Such capabilities are demonstrated throughwalking on the ground and grasping objects using the same leg mechanism.Item Open Access Efficient representation, measurement, and recovery of spatial and social networks(Colorado State University. Libraries, 2021) Mahindre, Gunjan S., author; Jayasumana, Anura, advisor; Paffenroth, Randy, committee member; Maciejewski, Anthony, committee member; Kirby, Michael, committee memberTo view the abstract, please see the full text of the document.Item Embargo Energy-aware workload management for geographically distributed data centers(Colorado State University. Libraries, 2023) Hogade, Ninad, author; Pasricha, Sudeep, advisor; Siegel, Howard Jay, committee member; Maciejewski, Anthony, committee member; Anderson, Chuck, committee memberCloud service providers are distributing data centers globally to reduce operating costs while also improving the quality of service by using intelligent cloud management strategies. The development of time-of-use electricity pricing and renewable energy source models has provided the means to reduce high cloud operating costs through intelligent geographical workload distribution. However, neglecting essential considerations such as data center cooling power, interference effects from workload co-location in servers, net-metering, peak demand pricing of electricity, data transfer costs, and data center queueing delay has led to sub-optimal results in prior work because these factors have a significant impact on cloud operating costs, performance, and carbon emissions. This dissertation presents a series of critical research studies addressing the vital issues of energy efficiency, carbon emissions reductions, and operating cost optimization in geographically distributed data centers. It scrutinizes different approaches to workload management, considering the diverse, dynamic, and complex nature of these environments. Starting from an exploration of energy cost minimization through sophisticated workload management techniques, the research extends to integrate network awareness into the problem, acknowledging data transfer costs and queuing delays. These works employ mathematical and game theoretic optimization to find effective solutions. Subsequently, a comprehensive survey of state-of-the-art Machine Learning (ML) techniques utilized in cloud management is discussed. Then, the dissertation traverses into the realm of Deep Reinforcement Learning (DRL) based optimization for efficient management of cloud resources and workloads. Finally, the study culminates in a novel game-theoretic DRL method, incorporating non-cooperative game theory principles to optimize the distribution of AI workloads, considering energy costs, data transfer costs, and carbon footprints. The dissertation holds significant implications for sustainable and cost-effective cloud data center workload management.Item Open Access Looking under the hood: visualizing what LSTMs learn(Colorado State University. Libraries, 2019) Patil, Dhruva, author; Draper, Bruce, advisor; Beveridge, J. Ross, committee member; Maciejewski, Anthony, committee memberRecurrent Neural Networks (RNNs) such as Long Short Term Memory (LSTM) and Gated Recurrent Units (GRUs) have been successful in many applications involving sequential data. The success of these models lies in the complex feature representations they learn from the training data. One criteria to trust the model is its validation accuracy. However, this can lead to surprises when the network learns properties of the input data, different from what the designer intended and/or the user assumes. As a result, we lack confidence in even high-performing networks when they are deployed in applications with novel input data, or where the cost of failure is very high. Thus understanding and visualizing what recurrent networks have learned becomes essential. Visualizations of RNN models are better established in the field of natural language processing than in computer vision. This work presents visualizations of what recurrent networks, particularly LSTMs, learn in the domain of action recognition, where the inputs are sequences of 3D human poses, or skeletons. The goal of the thesis is to understand the properties learned by a network with regard to an input action sequence, and how it will generalize to novel inputs. This thesis presents two methods for visualizing concepts learned by RNNs in the domain of action recognition, providing an independent insight into the working of the recognition model. The first visualization method shows the sensitivity of joints over time in a video sequence. The second visualization method generates synthetic videos that maximize the responses of a class label or hidden unit within a set of known anatomical constraints. These techniques are combined in a visualization tool called SkeletonVis to help developers and users gain insights into models embedded in RNNs for action recognition. We present case studies on NTU-RGBD, a popular data set for action recognition, to reveal properties learnt by a trained LSTM network.Item Open Access Modeling of twisted and coiled artificial muscle for actuation and self-sensing(Colorado State University. Libraries, 2018) Abbas, Ali, author; Zhao, Jianguo, advisor; Bradley, Thomas, committee member; Maciejewski, Anthony, committee memberSoft robots are a new type of robots with deformable bodies and muscle-like actuations, which are fundamentally different from traditional robots with rigid links and motor-based actuators. Owing to their elasticity, soft robots outperform rigid ones in safety, maneuverability, and adaptability. With their advantages, many soft robots have been developed for manipulation and locomotion in recent years. Nevertheless, two issues prevent the wide applications of developed soft robots: cumbersome actuation methods (e.g., pneumatics) and limited sensing capability to feedback the robot's shape. To address these two issues, this thesis leverages a recently discovered twisted and coiled artificial muscle for soft robots. This artificial muscle can generate large force and displacement; moreover, we recently found that it has self-sensing capability, i.e., its electrical resistance will increase if the muscle is elongated by an external force. With the dual actuation and self-sensing capability, we expect to accomplish closed-loop control of soft robots for precise motion without external sensors, potentially solving the two issues for existing soft robots. This thesis will focus on three aspects for the twisted and coiled artificial muscle. First, we model the actuation from a physics perspective. Such a model utilizes parameters related to the working principle and material properties of the actuator, eliminating the requirements for tedious system identifications. Experiments are conducted to verify the proposed model, and the results demonstrate that the proposed model can predict the static performance and dynamic response for the muscle. Second, we test and model the sensing capability of the artificial muscle. Specifically, we establish a physics-based model to predict the external force and the displacement if the resistance is given and experimentally validate its correctness. Third, we apply the actuation and sensing of the artificial muscle to soft robots. To demonstrate we can leverage the muscle to actuate soft robots, we fabricate a soft manipulator with multiple muscles as well as a robotic fish tail. To demonstrate the sensing capability, we embed the muscle into soft materials and successfully measure two curvatures of a two-segment soft robot. Based on the work presented in this thesis, our future work will integrate the actuation and sensing capability of the twisted and coiled artificial muscle to enable closed-loop shape control of soft robots.Item Open Access RIDMBC for object recognition using convolutional neural networks(Colorado State University. Libraries, 2016) Agnihotri, Nikhil, author; Draper, Bruce A., advisor; Beveridge, Ross, advisor; Maciejewski, Anthony, committee memberTwo trending techniques that are making advances in computer vision research are Convolutional Neural Networks and Visual Hashing. The goal of this paper is to analyze how these two interact in the broad domain of objects. Deep neural nets have proved to broadly represent image features, and binary codes have proved to be a powerful way to represent the intrinsic nature of image content in a compact way. Our research explores what kind of information is contained in feature vectors obtained from deep neural nets and what infor- mation can be binarized, in the context of object recognition. We also try to optimize the length of binary codes and select subsets of bit vectors to represent images so as to obtain the best classification results, while trying to bring down computational cost.Item Open Access Robust and secure resource management for automotive cyber-physical systems(Colorado State University. Libraries, 2022) Kukkala, Vipin Kumar, author; Pasricha, Sudeep, advisor; Maciejewski, Anthony, committee member; Pezeshki, Ali, committee member; Bradley, Thomas, committee memberModern vehicles are examples of complex cyber-physical systems with tens to hundreds of interconnected Electronic Control Units (ECUs) that manage various vehicular subsystems. With the shift towards autonomous driving, emerging vehicles are being characterized by an increase in the number of hardware ECUs, greater complexity of applications (software), and more sophisticated in-vehicle networks. These advances have resulted in numerous challenges that impact the reliability, security, and real-time performance of these emerging automotive systems. Some of the challenges include coping with computation and communication uncertainties (e.g., jitter), developing robust control software, detecting cyber-attacks, ensuring data integrity, and enabling confidentiality during communication. However, solutions to overcome these challenges incur additional overhead, which can catastrophically delay the execution of real-time automotive tasks and message transfers. Hence, there is a need for a holistic approach to a system-level solution for resource management in automotive cyber-physical systems that enables robust and secure automotive system design while satisfying a diverse set of system-wide constraints. ECUs in vehicles today run a variety of automotive applications ranging from simple vehicle window control to highly complex Advanced Driver Assistance System (ADAS) applications. The aggressive attempts of automakers to make vehicles fully autonomous have increased the complexity and data rate requirements of applications and further led to the adoption of advanced artificial intelligence (AI) based techniques for improved perception and control. Additionally, modern vehicles are becoming increasingly connected with various external systems to realize more robust vehicle autonomy. These paradigm shifts have resulted in significant overheads in resource constrained ECUs and increased the complexity of the overall automotive system (including heterogeneous ECUs, network architectures, communication protocols, and applications), which has severe performance and safety implications on modern vehicles. The increased complexity of automotive systems introduces several computation and communication uncertainties in automotive subsystems that can cause delays in applications and messages, resulting in missed real-time deadlines. Missing deadlines for safety-critical automotive applications can be catastrophic, and this problem will be further aggravated in the case of future autonomous vehicles. Additionally, due to the harsh operating conditions (such as high temperatures, vibrations, and electromagnetic interference (EMI)) of automotive embedded systems, there is a significant risk to the integrity of the data that is exchanged between ECUs which can lead to faulty vehicle control. These challenges demand a more reliable design of automotive systems that is resilient to uncertainties and supports data integrity goals. Additionally, the increased connectivity of modern vehicles has made them highly vulnerable to various kinds of sophisticated security attacks. Hence, it is also vital to ensure the security of automotive systems, and it will become crucial as connected and autonomous vehicles become more ubiquitous. However, imposing security mechanisms on the resource constrained automotive systems can result in additional computation and communication overhead, potentially leading to further missed deadlines. Therefore, it is crucial to design techniques that incur very minimal overhead (lightweight) when trying to achieve the above-mentioned goals and ensure the real-time performance of the system. We address these issues by designing a holistic resource management framework called ROSETTA that enables robust and secure automotive cyber-physical system design while satisfying a diverse set of constraints related to reliability, security, real-time performance, and energy consumption. To achieve reliability goals, we have developed several techniques for reliability-aware scheduling and multi-level monitoring of signal integrity. To achieve security objectives, we have proposed a lightweight security framework that provides confidentiality and authenticity while meeting both security and real-time constraints. We have also introduced multiple deep learning based intrusion detection systems (IDS) to monitor and detect cyber-attacks in the in-vehicle network. Lastly, we have introduced novel techniques for jitter management and security management and deployed lightweight IDSs on resource constrained automotive ECUs while ensuring the real-time performance of the automotive systems.Item Open Access Secure, accurate, real-time, and heterogeneity-resilient indoor localization with smartphones(Colorado State University. Libraries, 2022) Tiku, Saideep, author; Pasricha, Sudeep, advisor; Pallickara, Shrideep, committee member; Maciejewski, Anthony, committee member; Siegel, H. J., committee memberThe advent of the Global Positioning System (GPS) reformed the global transportation industry and allowed vehicles to not only localize themselves but also to navigate reliably and in a secure manner across the world at high speeds. Today, indoor localization is an emerging IoT domain that is poised to reinvent the way we navigate within buildings and subterranean locales, with many benefits, e.g., directing emergency response services after a 911 call to a precise location (with sub-meter accuracy) inside a building, accurate tracking of equipment and inventory in hospitals, factories, and warehouses, etc. While GPS is the de-facto solution for outdoor positioning with a clear sky view, there is no prevailing technology for GPS-deprived areas, including dense city centers, urban canyons, and inside buildings and other covered structures, where GPS signals are severely attenuated or totally blocked, and affected by multipath interference. Thus, very different solutions are needed to support localization in indoor locales. Popular solutions for indoor positioning with high accuracy leverage wireless radio signals, such as WiFi, Bluetooth ultra-wideband (UWB), etc. Due to the existing widespread deployment of WiFi access points (WAPs) in most indoor locales, using WiFi for indoor localization can lead to low-cost solutions. Many localization algorithms that utilize these wireless signals have been proposed, e.g., based on the principles of proximity, trilateration, triangulation, and fingerprinting. Studies have shown that fingerprinting-based algorithms deliver higher accuracy, without stringent synchronization or line-of-sight requirements and enable greater error resilience in the presence of frequently encountered multipath signal interference effects, than other alternatives. A fingerprinting-based approach for indoor localization has two phases. In an offline phase, location-tagged wireless signal signatures, i.e., fingerprints, at known indoor locations are captured along a path and stored in a database. Each fingerprint in the database consists of a location and wireless signal characteristics, e.g., received signal strength (RSSI; which varies as a function of distance from the WAP), from visible WAPs at that location. This phase requires great manual effort of collecting several fingerprints at each location and comes at considerable cost. In the online phase, the observed RSS on the user's mobile device is used to query the fingerprint database and determine location (potentially after some interpolation). Such WiFi-based fingerprinting is a promising building block for low-cost indoor localization with mobile devices. Unfortunately, there are many unaddressed challenges before a viable WiFi fingerprinting based solution can be realized: (i) the algorithms used for the matching of fingerprints in the online phase have a major impact on accuracy, however the limited CPU/memory/battery resources in mobile devices requires careful algorithm design and deployment that can trade-off accuracy, energy-efficiency, and performance (localization decision latency); (ii) the diversity of mobile devices poses another challenge as smartphones from different vendors may have varying device characteristics leading to different fingerprints being captured at the same location; (iii) security vulnerabilities due to unintentional or intentional WiFi jamming and spoofing attacks can create significant errors which must be overcome; and (iv) short-term and long-term variations in WAP power levels and the indoor environments (e.g., adding/moving furniture, equipment, changes in density of people) can also introduce errors during location estimation, that often corrected by the expensive collecting new fingerprints. In this dissertation, we propose a new real-time machine learning based framework called SARTHI that addresses all of the abovementioned key challenges towards realizing a viable indoor localization solution with smart mobile devices. To enable energy-efficient enhancements in localization accuracy, SARTHI includes lightweight yet powerful machine learning algorithms with a focus on achieving a balance between battery life and response time. To enable device heterogeneity resilience, we analyzed and identified device diversity invariant pattern matching metrics that can be incorporated into a variety of machine learning based indoor localization frameworks. SARTHI also addresses the challenges associated with the security of fingerprinting-based indoor localization frameworks in the presence of spoofing and jamming attacks. This is achieved by devising a novel methodology for training and deploying deep-learning algorithms that are specifically designed to be resilient to the vulnerabilities associated with intentional power level variation-based attacks. Finally, SARTHI addresses the challenges associated with short-term and long-term variations in WiFi fingerprints using novel low-overhead relativistic learning-based deep-learning algorithms that can deliver high-accuracy while simultaneously minimizing the fingerprint collection effort in the offline phase.Item Open Access Soft and shape morphing robots driven by twisted-and-coiled actuators(Colorado State University. Libraries, 2022) Sun, Jiefeng, author; Zhao, Jianguo, advisor; Maciejewski, Anthony, committee member; Gao, Xinfeng, committee member; Yourdkhani, Mostafa, committee memberSoft robots are a new type of robot with deformable bodies and muscle-like actuation, which are fundamentally different from traditional robots with rigid links and motor-based actuators. Owing to their elasticity, soft robots outperform rigid ones in safety, maneuverability, and adaptability. With their advantages, many soft robots have been developed for manipulation and locomotion in recent years. To enable their unique capabilities, soft robots require a key component—the actuator. Many different actuators have been used, including the conventional pneumatic-driven and cable-driven methods, as well as several novel approaches proposed recently such as combustion, dielectric elastomers, chemical reactions, liquid–vapor transition, liquid crystal elastomer, and shape memory alloy. Besides existing actuation approaches, another promising actuator for soft robots is the twisted-and-coiled actuator (TCA). Compared with existing actuation methods, TCAs exhibit several unique characteristics: like large energy density and being directly actuated by electricity with a small voltage. All of these characteristics will potentially enable small-scale and untethered soft robots that in general are difficult to be accomplished by pneumatic and tendon-driven methods. Further, unlike shape memory alloys, TCAs are intrinsically soft, making it possible to embed them in any shape inside a soft body to generate versatile motion. To better actuate soft robots with TCAs, we introduce a novel fabrication technique of contraction TCAs that will have uniform initial gaps between neighboring coils. In this case, they can contract larger than 48% without a preload, termed free stroke. We also characterize such a TCA and compare it with self-coiled TCAs. Besides the free stroke property, the TCA can also be directly used as a sensor that provides its displacement information. To better design, optimize, and control TCAs for various applications, we developed a physics-based model based on TCAs' physical parameters as opposed to system identification methods, since such physics-based models are expected to be a general model for different types of TCAs (self-coiled, free-stroke, conical) We demonstrate soft robots with programmable motions by placing TCAs in different shapes inside a soft body. Specifically, we embed TCAs in a curved U shape, a helical shape, and straight shapes in parallel to enable three different motions: two-dimensional bending, twisting, and three-dimensional bending. We also combine the three motions to demonstrate a completely soft robotic arm that mimics a human forearm. A model is also developed to simulate the TCA-driven soft robots. The framework can model 1) the complicated routes of multiple TCAs in a soft body and 2) the coupling effect between the soft body and the TCAs during their actuation process. When not actuated, a TCA in the soft body is an antagonistic elastic element that restrains the magnitude of the motion and increases the stiffness of the robot. By stacking several modules together, we simulate the sequential motion of a soft robotics arm with three-dimensional bending, twisting, and grasping motion. The presented modeling and simulation approach will facilitate the design, optimization, and control of soft robots driven by TCAs or other types of artificial muscles. Finally, we design shape morphing robots that can morph the shape of their bodies to adapt to a different environment. These robots can be built with shape-morphing modules. A shape-morphing module has a variable stiffness element that allows it to switch between soft and rigid states. While it is in a soft state, it can morph to different configurations driven by TCAs. We demonstrate robots built with these modules can morph to different shapes that facilitate grasping and locomotion.Item Open Access Spatiotemporal anomaly detection: streaming architecture and algorithms(Colorado State University. Libraries, 2020) Siegel, Barry W., author; Labadie, John, advisor; Chong, Edwin, committee member; Maciejewski, Anthony, committee member; Young, Peter, committee memberAnomaly detection is the science of identifying one or more rare or unexplainable samples or events in a dataset or data stream. The field of anomaly detection has been extensively studied by mathematicians, statisticians, economists, engineers, and computer scientists. One open research question remains the design of distributed cloud-based architectures and algorithms that can accurately identify anomalies in previously unseen, unlabeled streaming, multivariate spatiotemporal data. With streaming data, time is of the essence, and insights are perishable. Real-world streaming spatiotemporal data originate from many sources, including mobile phones, supervisory control and data acquisition enabled (SCADA) devices, the internet-of-things (IoT), distributed sensor networks, and social media. Baseline experiments are performed on four (4) non-streaming, static anomaly detection multivariate datasets using unsupervised offline traditional machine learning (TML), and unsupervised neural network techniques. Multiple architectures, including autoencoders, generative adversarial networks, convolutional networks, and recurrent networks, are adapted for experimentation. Extensive experimentation demonstrates that neural networks produce superior detection accuracy over TML techniques. These same neural network architectures can be extended to process unlabeled spatiotemporal streaming using online learning. Space and time relationships are further exploited to provide additional insights and increased anomaly detection accuracy. A novel domain-independent architecture and set of algorithms called the Spatiotemporal Anomaly Detection Environment (STADE) is formulated. STADE is based on federated learning architecture. STADE streaming algorithms are based on a geographically unique, persistently executing neural networks using online stochastic gradient descent (SGD). STADE is designed to be pluggable, meaning that alternative algorithms may be substituted or combined to form an ensemble. STADE incorporates a Stream Anomaly Detector (SAD) and a Federated Anomaly Detector (FAD). The SAD executes at multiple locations on streaming data, while the FAD executes at a single server and identifies global patterns and relationships among the site anomalies. Each STADE site streams anomaly scores to the centralized FAD server for further spatiotemporal dependency analysis and logging. The FAD is based on recent advances in DNN-based federated learning. A STADE testbed is implemented to facilitate globally distributed experimentation using low-cost, commercial cloud infrastructure provided by Microsoft™. STADE testbed sites are situated in the cloud within each continent: Africa, Asia, Australia, Europe, North America, and South America. Communication occurs over the commercial internet. Three STADE case studies are investigated. The first case study processes commercial air traffic flows, the second case study processes global earthquake measurements, and the third case study processes social media (i.e., Twitter™) feeds. These case studies confirm that STADE is a viable architecture for the near real-time identification of anomalies in streaming data originating from (possibly) computationally disadvantaged, geographically dispersed sites. Moreover, the addition of the FAD provides enhanced anomaly detection capability. Since STADE is domain-independent, these findings can be easily extended to additional application domains and use cases.Item Open Access Toolless out of build plane manufacturing of intricate continuous fiber reinforced thermoplastic composites with a 3D printing system(Colorado State University. Libraries, 2019) Bourgeois, Mark Elliott, author; Radford, Donald, advisor; Ma, Kaka, committee member; Maciejewski, Anthony, committee memberContinuous fiber reinforced composite materials are manufactured using a variety of techniques ranging from manual layup to highly automated tape and fiber placement, yet all of the processes require significant tooling to act as a form which gives the composite the desired shape until processing is complete. Once processed and rigid, the composite is removed from the tooling and the tooling is, usually, then prepared and another composite component shaped on the tool. Manufacturing on such tooling has the advantage of offering a repeatable shape in a large batch production of fiber reinforced composite parts; however, the tooling itself can be a significant time to manufacture and cost challenge. It may take a large volume of composite parts to effectively amortize the cost of the tooling, which has a finite service life. Further, once the tooling is produced, making geometry changes during a production cycle is almost impossible. Geometry changes need either remanufacture of the tooling or the development of completely new tooling sets. Thus, technologies which could reduce the required tooling for composites production are highly desirable. With the advent of additive manufacture, it has become commonplace to expect the development of components of very complex geometry built from a simple surface. However, unlike continuous fiber reinforced composites, these complex geometry 3D printed components have material properties which are, for the most part, non-directional. While fiber reinforced composites are produced in a layer-by-layer additive fashion, the key to the performance of this material family is the positioning and orientation of the continuous fiber over complex contours, which has resulted in a need for substantial tooling. Thus, if concepts related to 3D printing could be mapped into the continuous fiber reinforced composite manufacturing space, the potential may exist for a radical reduction in the amount of tooling required and a corresponding increase in the flexibility of manufacture. The current research effort implements concepts common to 3D printing to investigate an approach to producing continuous fiber reinforced structures which require no tooling. Sandwich panels are commonly used as structure based on fiber reinforced composites, with the goal of high flexural stiffness and low mass. It is most common to separate two high performance composite laminates (facesheets) with a low-density core material, generally in the form of a foam of honeycomb. A recent concept has been to replace these traditional core materials with fiber reinforced truss-like structures, with the goal of further reducing mass; however, a manufacturable solution for these truss core sandwich panels has not been developed and those processes that do exist are tooling intensive. In this work, a system was developed and demonstrated that can radically reduce the amount of tooling required for truss core sandwich panels. Pyramidal truss core sandwich panels were manufactured to test the positional fidelity of out of build plane, unsupported space manufacturing. Laminates with different lamina counts were manufactured on a substrate and in unsupported space and tested for consolidation quality. Lap shear specimens were manufactured on a substrate and in unsupported space and tested for interlaminar bond quality. Individual continuous fiber reinforced composite strand specimens were manufactured in unsupported space at varying temperatures and tested for stiffness. These truss core panels, manufactured without tooling, were compared to similar truss core panels produced by more traditional techniques. The outcome of the research performed indicates that structures could be manufactured, unsupported, in free space with good precision. The void content of laminates manufactured in unsupported space decreased by 15% as the laminate was built up while the laminate manufactured on the substrate had no significant change in void content. Unidirectional laminates placed in space showed no statistical difference in strength when compared to laminates placed on a substrate. Crossply laminates had a 33% reduction in strength compared to similar laminates placed on a substrate. Composite truss core sandwich panels manufactured with the system developed in this work were more precise than composite truss core sandwich panels manufactured with compression molding and heat fusion bonding. Increasing the placement temperature of continuous fiber reinforced thermoplastic strands increases the quality of the strand by up to 44%. Improvements to the MAGIC system have increased the composite quality by 25%. Thus, manufacturing techniques were implemented to place fiber not only within the build plane, X-Y, but also to place continuous fiber out of the build plane, X Y Z. Intricate continuous fiber reinforced thermoplastic composites were manufactured without the use of tooling. While the composites produced with the new system were less stiff than composites made with compression molding further improvements to the manufacturing system have closed the stiffness gap between the two manufacturing methods.