DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Solutions and analysis of multivariate polynomial systems for geolocation and initial orbit determination
No full textMay 2023School of EngineeringWith the birth of Space Situational Awareness from the launch of the Russian satellite, Sputnik I, humanity has found it necessary to develop methods to determine the navigational state of man-made objects orbiting the Earth for avoidance and tracking purposes. With the ever-increasing number of unknown objects, both ground-based and orbiting, the necessity of these methods has never been greater. This work seeks to further the fields of Spacecraft Navigation and Space Situation Awareness with the development of three novel solutions to the geolocation and initial orbit determination problems. The first method is a geolocation technique utilizing both time-based and frequency-based measurements from the signal of a ground-based Radio Frequency (RF) transmitter. Our second method derives two initial orbit determination techniques using concurrent TDOA and range-rate measurements from the signal of an orbiting RF transmitter. The final method discussed in this work is an initial relative orbit determination method using range-rate measurements and the linearized Clohessy Wiltshire dynamics. Each method derives its solution from a polynomial system solved using an algebraic geometry technique called homotopy continuation theory. Each method is verified using simulation results for multiple scenarios.Ph
Stochastic modeling of intracellular transport in neurons
No full textAugust 2019School of ScienceA cell is not just a cell. It’s a factory that is manufacturing, growing, signaling, communicating, transporting, dividing, all at the same time. Molecular motors are the powerful agents that perform the long distance transport in eukaryotic cells of humans and other animals. Understanding of these transport processes is a crucial step towards developing further insights into many neuro-degenerative disorders that have plagued our species. We present a mathematical framework to analyze the intracellular transport inside a neuron. We study the transport on a parallel arrangement of microtubules inside the axon (axonal transport), as well as various tangled networks of microtubules inside the soma (somatic transport). For the former, our model captures the spatial dynamics and interactions of a motor and cargo particles, and the mean attachment time of the motor to a microtubule is computed. For the latter, we analyze the effects on the transport of particle switching at the microtubule intersections. In all cases, we have obtained the effective velocity and diffusion coefficient for the transport at the cellular scale. To validate the theoretical results for the motor attachment time, we also present a custom-built numerical scheme to efficiently simulate the mean first passage time to a small target in a multidimensional domain.Ph
Atomic and electronic structures of extended defects, disorder, and reactions in solids
No full textDecember 2014School of ScienceAs the physicist Sir Charles Frank said, `Crystals are like people: it is the defects in them that make them interesting.' Although tiny in size, the defects in the microscopic scale would determine the macroscale functionality of materials. For the purpose to design, functionalize and engineer materials, the fundamental understanding and tailoring defects and other disorder features are paramount. This thesis focuses on the extended defects studies in solids, through first-principles calculations and other theoretical methods, concerning their atomic and electronic related properties. The thesis is organized in three parts: Part I: Extended defects in inorganic materials; PART II: Disorders in organic conducting polymers/molecule crystals; PART III: Nanosize effect on the defect-mediated reactant diffusion for catalysis reaction. In Part I, we are mainly focusing on the extended defects in inorganic materials, with their forms as dislocations, grain boundaries, and etc. This part starts with the general introduction to extended defects in Chapter 1, followed by three different case studies from Chapter 2 to 4. Chapter 2 presents the multi-scale simulation of small-angle grain boundaries in Si. As electrically-benign, low-cost, near-single crystalline semiconductors is crucial for large-scale applications, understanding the electrical behavior of small-angle grain-boundaries (GBs), consisting of array of dislocations, becomes important. However, a quantitative microscopic theory is still prohibitively difficult. Here, by developing a multiscale approach combining Monte-Carlo simulation, elasticity theory, first-principles calculation, and grand-canonical statistical modeling, we quantitatively explain the recent experimental observation showing the disappearance of hole transport barrier in poly-Si when the GB angle is below a critical value of a few degrees. It reveals the microscopic origin for the observation as a transition from electrically harmful dislocations to electrically benign dislocations. Chapter 3 presents our developed modular approach for calculations of an epitaxial interface formed between materials with a large misfit (f ≥ 10%). Two different types of epitaxial interfaces, metal-semiconductor with covalent bonds at interfaces and semiconductor-semiconductor with Coulombic interactions at interfaces, are studied through this method. Both of them demonstrate good agreements with experimental results. Via the electron localization function (ELF), we observe and quantify the directional bonding density ρb at the interface. For the metal-semiconductor interfaces (Al/Si, Cu/Si), the ρb is found responsible for the experimentally observed epitaxial relationships. For the semiconductor-semiconductor interface formed between CdTe film and As-passivated Ge with no dangling bonds on the Ge substrate surface, a weaker interface binding with ρb = 0 is obtained. Chapter 4 goes beyond the extended defects in three-dimensional (3D) materials, focusing on the extended defects in 2D heterostructures of graphene and hexagonal boron nitride (h-BN) monolayers. Guided by the Clar's sextet rule, the stable heterostructures at different misorientaion angles are modeled and calculated. Based on these structures, the Schottky barrier height (SBH) for 2D graphene/h-BN lateral heterostructures is computed using first-principles methods. The calculated SBH is remarkably insensitive to the misorientation angle between graphene and h-BN and close to the ideal SBH that follows the Schottky-Mott limit. This unexpected result can be quantitatively reproduced by an infinitely long dipole line model developed for 2D systems. The quickly decaying of interface dipole induced potential difference is responsible for the universal SBH. The present results raise the possibility of fabricating high-quality transport devices using current graphene/h-BN heterostructures and establish the principle that any 2D heterostructure follows the Schottky-Mott limit when the device size is large than several nanometers. In Part II, we discuss disorders in organic materials, specifically, the conducting polymers/molecule crystals. Compared with inorganic crystals, organic polymers or molecules crystals have higher degree of freedom and thus more room for disorders. Furthermore, because of the technical difficulty of simulating the weak van der Waals (vdW) interaction in these systems, this disorder problem was hard to be accurately described before. In this part, we will introduce our developed DFT+LAP method that is capable to describe vdW interactions accurately. Two prototype conducting inorganic materials P3HT and DNTT(-C10) are studied. In Chapter 5, the crystal structure of poly(3-hexylthiophene) (P3HT) has been studied by first-principles calculations based on density functional theory. The generalized gradient approximation is employed and van der Waals interactions are treated accurately by the recently proposed local atomic potential (LAP) approach. A variety of different models were tested, and the model having the lowest energy is a non-interdigitated structure having an orthorhombic cell with a = 17.2 Å, b = 7.7 Å, and c = 7.8 Å, where a, b, and c are the lengths of the lattice vectors perpendicular to the lamallae, in the π-π stacking direction, and along the thiophene backbone, respectively. These values are in reasonably good agreement with experiment. The P3HT polymer is not invariant under inversion and therefore exhibits directionality. Our calculations suggest that a likely structural defect occurring in P3HT is one in which one of the polymer backbones within a lamella runs in the direction opposite to the majority. Such defects may form in the process of self-assembly of the non-interdigitated lamellae and may be an important source of π-π stacking disorder. A possible explanation for a recently observed structural phase transition in polythiophene is proposed. In Chapter 6, we present a comparison based on first-principles calculations of the electronic structure of non-alkylated and alkylated dinaphtho[2,3-b:2′,3′-f]thieno[3,2-b]thiophene: DNTT and DNTT-C10. The calculations show that the addition of alkyl chains decreases intermolecular distances, in agreement with experiments. Calculations indicate that effective masses are reduced by the addition of alkyl chains, and within a simple deformation potential model, this translates into higher mobility for DNTT-C10. The shorter intermolecular distances found in DNTT-C10 are attributed to van der Waals interactions between alkyl chains. In the final part of this thesis (Part III), we study the interesting "side-effect" of disorders. For example, how would the nano-size effect in reactant affect the catalytic reaction that is caused by catalysts? It is found that chemical reactions involve competing pathways. Here, two pathways for H desorption from MgH2 are identified by first-principles calculations: one involves H diffusion in bulk Mg, while the other involves H vacancy diffusion at the MgH2 surface. The latter is sensitive to the size of the reactant MgH2 and self-terminates in bulk material as dehydrogenation eventually eliminates exposed MgH2. However, this surface vacancy pathway can maximize the catalyst effect of Pd by decoupling the kinetics of H desorption from that of H diffusion. When the surface-to-bulk ratio is large as in the case of MgH2 nanostructures, H desorption will take place primarily via the low-barrier surface vacancy pathway. Our picture attributes the experimentally observed size effect of MgH2 on H desorption, beyond the quantum size regime, to a synergy between the nanosize of the reactant and the catalyst.Ph
Molecular determinants of trade-offs between antibody colloidal stability and non-specificity
No full textMay 2021School of ScienceThe widespread interest in antibody therapeutics has led to much focus on identifying antibody candidates with favorable developability properties. In particular, there is broad interest in identifying antibody candidates with highly repulsive antibody self-interactions in standard formulation conditions (e.g., low ionic strength buffers at pH 5-6) for high solubility and low viscosity in concentrated antibody formulations. Likewise, there is also broad interest in identifying antibody candidates with low levels of non-specific interactions in physiological conditions (PBS, pH 7.4) to promote favorable pharmacokinetic properties by minimizing non-specific binding to cells and tissues. Here we demonstrate that these two objectives are at natural odds with each other based on our analysis of 42 IgG1 variants with variable fragments (Fv) from four clinical-stage antibodies. Notably, we find that antibodies with the most repulsive self-interactions in standard formulation conditions have the highest levels of non-specific binding in physiological conditions. Conversely, antibodies with the lowest levels of non-specific binding in physiochemical conditions display the highest levels of self-association in standard formulation conditions. These results are largely explained by the fact that the antibody net charges and isoelectric points are most strongly correlated with both types of antibody interactions, as antibodies with increased positive charge displayed decreased antibody self-association in formulation conditions and increased non-specific interactions in physiological conditions. IgG1s with isoelectric points between 8-8.5, and Fv isoelectric points between 7.5-9, generally displayed the best combination of properties, as they generally possessed strongly repulsive self-interactions and low-to-moderate levels of non-specific interactions. We expect these findings will improve the identification and engineering of antibody candidates with drug-like biophysical properties.Ph
Hierarchical 3d structures for bioengineering applications
No full textDecember 2022School of EngineeringDevices and scaffolds made from soft polymeric materials are being developed for various applications like filtration, catalysis, tissue engineering, and regenerative medicine. To cater to different fields, these scaffolds are designed to be of varying length scales and geometries (flat/tubular) and manufactured using diverse materials. They serve as platforms to replicate the chemical/physical processes usually carried out by natural biological systems. One of the biological system's most critical aspects is its hierarchical self-repeating nature across size scales. This research aims to develop manufacturing processes for fabricating soft polymeric scaffolds that display hierarchy across different length scales. The first type of hierarchy replicated in the polymeric scaffold is based on the morphological hierarchy of branched tubular structures that constitute the circulatory system in the human body. The second type of hierarchy is the hierarchy of the materials that comprise the microstructure of the polymeric scaffold. The first manufacturing process developed as a part of this research seeks to fabricate free-standing tubular structures that display (i) the hierarchy of sizes (internal tube diameters) and (ii) the interconnectedness of the different-sized tubes in a branching pattern. This was accomplished by developing a hybrid manufacturing process that combines alginate electrodeposition with fiber electrospinning. Alginate electrodeposition was used to fabricate tubular structures with internal diameters at the micro-scale (~100 µm). The conventional electrodeposition process was modified by incorporating a moving electrode to be able to fabricate branched/bifurcated tubular structures. Branched tubular structures were manufactured through fiber electrospinning at the macro-scale (~10mm. These tubular constructs at the two extreme ends of the scale were joined together using the electrodeposition process at a common size scale. The final structure obtained displayed a hierarchy of tube diameters. Hierarchical prototypes were fabricated with differing branching architectures and could support flow similar to the physiological circulatory system.
The second manufacturing process seeks to replicate the hierarchical nature of the fibrous extracellular matrix (ECM) by combining two soft polymeric systems that are fibrous in nature and have different-sized fiber components, i.e., (i) Bacterial Cellulose (BC) fibers (~100 nm fiber diameter) and (ii) Polymeric nanofibers obtained through electrospinning (~1m fiber diameter). The integration of bacterial cellulose with the electrospun fibers was carried out using a bioreactor. The oscillatory motion bioreactor (OMB) added the BC-producing bacteria and the requisite nutrients into the electrospun fiber template in a periodic fashion. The frequency of the nutrient provision dictates the amount of BC fibers synthesized and the extent of integration with the electrospun fiber template. The resulting fibrous scaffolds showed a hierarchical microstructure with effective integration of electrospun nanofibers and BC fibers.
Finally, an exploratory manufacturing process was pursued to introduce hierarchical tubular channels into a bacterial cellulose matrix using biocompatible sacrificial material. Alginate tubes (500-800 m external diameter) were incorporated into BC culture media as a sacrificial material to introduce microchannel-shaped porosity into BC. When the tubes were washed away following BC synthesis, a microstructure with a hierarchy of pore sizes, i.e., tubular pores from alginate tubes and random pores present naturally in the BC scaffold, was achieved. The tubular pores could support fluid flow and be functional. In summary, the three manufacturing processes demonstrated in this dissertation are expected to benefit both the biomedical sector and applications requiring novel thin films involving bacterial cellulose structure.Ph
The virgo radial merger: a recent and substantial contribution to the local stellar halo
No full textAugust 2023School of ScienceThe Milky Way's (MW) stellar halo is built up from the debris of many disrupted dwarf galaxies. It is widely known that the MW's inner stellar halo contains a [Fe/H]-rich, radialized component, which contributes the majority of halo stars near the Sun, and is often referred to as the MW's "last major merger." The prevailing hypothesis for the origin of this component is that a sizeable dwarf galaxy collided with the MW proto-disk 8-11 Gyr ago, which is known as the \textit{Gaia}-Sausage/Enceladus (GSE) merger event. In this work we consider several collections of observed chemical and kinematic data for MW halo stars; our analyses of these data collectively come to the conclusion that the observed properties and structures in the MW stellar halo are inconsistent with a collision between the MW and a large dwarf galaxy 8-11 Gyr ago (as is proposed to explain the GSE). Instead, these observations are better explained as the result of a collision between a dwarf galaxy and the MW disk that occurred definitely less than 5 Gyr ago, and probably less than 3 Gyr ago, which we call the Virgo Radial Merger (VRM). We start by re-analyzing existing datasets of RR Lyrae (RRL) stars in the Virgo Overdensity (VOD) in the MW stellar halo. These stars originally had line-of-sight velocity measurements, but we are able to add proper motions to obtain their 3-dimensional velocities. The VOD contains a multitude of previously identified overlapping stellar streams and moving groups; we show that it is possible to generate the majority of the known structures using a simulation of a single dwarf galaxy passing through the center of the MW on a radial orbit (the VRM). This radial orbit gives the VRM debris a large range of energies and nearly zero Lz angular momentum. We potentially link the VRM with the Hercules-Aquila Cloud (HAC) and the Eridanus-Phoenix Overdensity (EPO) halo structures. The simulation of the VRM produces debris in the Solar neighborhood similar to the material associated with the Gaia-Enceladus/Gaia Sausage (GSE) merger event, despite only being integrated for 2 Gyr. As the GSE supposedly requires an age of 8-11 Gyr, resolving this discrepancy becomes the major focus of the remainder of this work. Radial merger events generate shells, which are spherically concentric distributions of stars that have been observed in other galaxies. Because we believe that the progenitor dwarf galaxy of the VRM collided with the MW on a radial orbit, we looked for the shells this collision was expected to produce and were able to identify the first shell structures every found in the MW. There are 2 shells in the VOD region and 2 shells in the HAC region, and we associate these structures with the VRM. The morphology of shell structure depends on how long it has been since the progenitor dwarf became unbound; eventually, the debris becomes indistinguishable from background stars. We analyze phase mixing in a collection of radial merger -body simulations, and find that shell structure similar to that observed in the MW disappears by 5 Gyr after collision with the Galactic center; this provides an upper limit on the time since the VRM progenitor collided with the MW disk. By placing test particles at the location of the MW shells and integrating their orbits backwards in time, we are able to identify the time when their progenitor dwarf galaxy became unbound. Our result indicates that the progenitor of the VRM passed through the Galactic center 2.7 +/- 0.2 Gyr ago. Based on the time of collision, it is possible that the VRM is related to the phenomenon that created phase-space spirals in the vertical motion of the disk, and could have caused a burst of star formation in the inner disk. Shells are formed from a subset of stars in a radial merger; however, if one looks at all stars in a radial merger in phase space (vr vs. r), the merger event is described by one or more parabolic ``caustics'', which can be thought of as a generalization of shells. We analyze the local phase space distribution in Gaia DR3 data, and recover the recently-identified phase-space caustics, which agree well with analytical models. The observed caustics have positive caustic velocities (the velocity of the structure at maximum Galactocentric radius), which support a more recent merger time, because phase-mixed debris produces chevron patterns with zero caustic velocity. Other works have not always noted the significance of this difference when comparing observed and simulated distributions of phase-space folds. We compare the observed phase space data to an analog GSE merger from the FIRE-2 Latte simulations. This simulated analog assembles a density profile similar to that observed in the MW much earlier after collision than previous simulations; a late assembly of a smooth density profile had previously been an argument for the large age of the GSE. We utilize a quantitative metric (causticality) that measures how phase-mixed a given distribution is, and find that the observed local phase-space distribution most closely matches the simulated data 1.5 Gyr after collision, and certainly not later than 3 Gyr after collision. This is further evidence that the progenitor of the ``last major merger'' did not collide with the MW proto-disk at early times, as is thought for the GSE, but instead collided with the MW disk within the last few Gyr. This is consistent with our previous collision time estimate for the VRM. After our assessments of the dynamics of radial merger events, we begin analyses of the chemical abundances of halo stars near the Sun. Because stars that are formed in different environments will have different chemical abundances, in principle these chemical abundances can separate the MW stellar halo into its distinct progenitor merger events. First, we use halo dwarf stars with photometrically determined metallicities that are located within 2 kpc of the Sun to identify local halo substructure. The kinematic properties of these stars are inconsistent with arising from a single, dominant radial merger event, such as the GSE. We find a number of distinct components in the stellar halo, including the VRM, and two new components which we name Nereus and Cronus. Nereus is non-rotating and has similar enegies to the VRM, but has lower [Fe/H]. Cronus has higher [Fe/H] than the VRM, and is co-rotating with the disk. We also identify the Nyx Stream, one or two smooth halo ``background'' components, and a disk contamination and/or in-situ halo component. We then analyze the chemical and kinematic properties of giant stars near the Sun, and we find that the chemical abundances and dynamics of this data are also inconsistent with a scenario in which the inner halo is primarily composed of debris from a single, massive, ancient merger event. The trends of chemical composition with energy in the data are opposite to expectations for a single massive, ancient merger event. Also, multiple chemical evolution paths with distinct dynamics are present. We find that the data is best fit by a model with four components. These components are the same as before: the VRM, Nereus, and Cronus, plus a previously-identified component named Thamnos. Nereus and Thamnos likely represent more than one accretion event because the chemical abundance distributions of their member stars contain many peaks. Because the local stellar halo contains multiple substructures, different popular methods of selecting GSE stars will actually select different mixtures of these substructures. We conclude with a novel picture of the MW's accretion history. We find it likely that the MW experienced one or more sizeable merger events early on in its history, which are located predominantly inside the Solar circle. The disequilibrium features in the MW stellar halo, including cloud overdensities, shells, and caustics in phase space, are all due to a much more recent merger event: the VRM. The progenitor of the VRM collided with the MW disk 2-3 Gyr ago, and produced a population of [Fe/H]-rich, [Alpha/Fe]-poor stars in the local stellar halo. These stars have a double-lobed velocity distribution that is responsible for the high-|vr| portions of the Gaia-Sausage velocity structure.Ph
Individual and team performance and behavioral adaptation in competitive adversarial games: an application to elite biathlon
No full textAugust 2023School of EngineeringThis dissertation develops and explores prospects for investigating human performance and adaptive behavior in competitive, individual- and team-based settings. It is motivated by recent work on theoretical underpinnings of performance and adaptation in teamwork, but also by a two-fold methodological challenge: how to instrument the environments in which teamwork takes place and how to leverage the resulting data to produce detailed descriptions and insightful analysis of task performance. The research presented here is organized in relation to a modified version of the Input-Mediator-Output (IMO) of Mathieu and colleagues, in which Inputs are the performers and task attributes, Mediators are the processes they engage in while on task, Outputs are the results of these processes by the performers in the task context, and Moderators are the situational factors of the scenario, which ultimately provide the feedback to support the development of teams and their members. Study 1 addresses prospects for capturing various IMO elements within team-based electronic gaming environments. Specifically, Study 1 yields recommendations for using open-source games to capture data on the Inputs of Team Composition and Working History, the Mediators of team Workflow and Communication, and the team Output of Performance. The approach augments the standard IMO framework by assessing the capacity of these games to capture data on different types of Tasks (e.g., Planing vs. Contests/Battles) within Environments that may be either Stable or Turbulent. Many of these themes are addressed in the real-world environment of elite biathlon in Studies 2 and 3. Elite biathlon––a sport that combines cross-country skiing and precision target shooting––is well suited to the study of human adaptation in competitive settings. Equally importantly, it can be used to address many aspects of the framework described in Study 1. In terms of the IMO Inputs, athletes can compete as individuals or as members of a variety of configurations of sub-teams within a national team (Composition), thus accruing experience working with each other (Working History). In terms of IMO Mediators, limited information on the state of the race is available on-course through a competitor's experience and team staff---though clearly this information is neither complete nor necessarily current. In terms of IMO Outputs, there are clearly defined and meaningful intermediate and final measures of performance. Finally, Task type can be defined as "Contest/Battle" while certain aspects of the Environment (such as weather) may be used to determine the extent to which there is stability or turbulence around the athletes. It should be noted that, due to longstanding interest within biathlon in advanced instrumentation, data are available to support analysis of individual- and team-level development at an appropriate scale over fairly long time horizons. Study 2 develops and illustrates the use of theoretically grounded measures of Mediator and Output of task performance at the process (i.e., the product of task actions) and outcome (i.e., intermediate and final standings) levels for the mixed relay discipline of biathlon. This exploratory study examines possible relationships and trade-offs among three measures of process performance: effectiveness, efficiency, and equality. Considering the underlying uncertainty associated with behavioral responses to changes in a system, entropy is used here to characterize behavioral change and adaptation over time. Accordingly, correlations across values of relative intermediate outcome performance variables are examined. The results suggest that process efficiency and equality, rather than effectiveness, introduce more variability in outcome effectiveness. Process effectiveness was highly correlated with and drove outcome effectiveness. The results suggest that future work should explore aspects of a competitive system that are related to the underlying contributors to efficiency and equality in order to understand adaptive behavior. Study 3 further explores Output performance by examining performance outcomes (i.e., intermediate and final standings of individual races) as a function of Input variables (individual, team, and task) and one salient Moderator factor of the environment (wind speed). The cyclic aspects of the IMO framework are considered by the time-series of competitor performance over the course of a race, and by modeling data over five consecutive seasons of elite biathlon. Building on the methodological perspective of Study 2, this study considers the extent to which behavior serves to reduce uncertainty. Adaptation here is framed as a deviation from baseline behavior and expressed in terms of entropy. The study results suggest that deviations from established baseline behavior differ by type of competition (and therefore task type), and that some competitors' substantial deviation across task types is suggestive of adaptive behavior. The overall contributions from this work include 1) the connection of the theories of team behavior to the domain of adversarial competitions, thus creating new empirical results; 2) an underlying theoretical framework to support empirical studies in performance and behavioral adaptation in individual and team competitive adversarial games; 3) suggestions of alternative research paradigms (open-source electronic game environments and complex secondary sporting data) for studying human performance; and 4) application of advanced analytic methods to complex secondary data in the target domain of biathlon. This work has also revealed various challenges and limitations associated with the marshaling of complex secondary data in the development of new theory, the exploration of team-based phenomena in real-world settings over time, and the need for further work in identifying opportunities for measurement of target phenomena.Ph
Fostering cooperative intersection crossing for connected and automated traffic on urban corridors
No full textMay 2023School of EngineeringConnected and automated vehicle (CAV) technologies can revolutionize transportation by eliminating human errors and enabling timely information sharing. The potential of CAVs to improve traffic mobility, safety, and sustainability is substantial, particularly at intersections that often represent significant bottlenecks of roadway networks. However, uncertain traffic flows and intense conflicting movements make vehicle crossing reliability a vital concern. Communication between CAVs and infrastructure can also substantially affect CAV crossing performance at intersections. Hence, ensuring CAV intersection crossing reliability is of utmost importance in this dynamic and complex environment.To tackle these issues, this dissertation aims to foster reliable cooperative intersection crossing for CAV traffic on urban corridors by advancing research in four interrelated areas: (1) Analyzing CAV traffic dynamic behavior focusing on CAV traffic oscillation under perturbation, (2) Investigating the communication side impact on the safety and stability of CAV platoons, (3) Enhancing CAV crossing reliability by adopting the notion of virtual platooning to develop an adaptive cooperative intersection crossing control, and (4) Leveraging the adaptive cooperative control to the corridor level to eliminate potential gridlock due to spatial constraints and traffic demand variations.
The findings of this dissertation enhance our understanding of CAV traffic behavior and the impact of vehicle communication on CAV safety and efficiency. The models and control principles developed in this dissertation can help in the design of more effective control strategies at isolated intersections and corridors, as well as inform the development of new approaches to address the challenges and opportunities presented by the rapid development and deployment of CAV technologies.Ph
Multivariate techniques for investigating complex biomedical challenges
No full textMay2021School of EngineeringWhile multivariate statistical techniques have found wide-spread use in many areas of science and engineering, their application in the health care sector is not nearly as wide-spread and usually limited to certain areas. This work makes use of state-of-the-art multivariate analysis techniques by applying them to three life science problems where such an application has not previously been reported. Specific emphasis is placed on the fact that early clinical trials may include a smaller number of participants and a larger number of variables. These statistical techniques are able to identify variables that can lead to better treatments and diagnosis that may not have been considered otherwise, because these techniques can sort through multiple combinations of variables to find a combination with the highest classification potential. This can either aid in the diagnosis by showing which measurements are significantly different between a control group and an affected group, or these significantly different measurements can point to interventions that may be needed to normalize the variable profiles of the affected group. The first contribution of this work focuses on the use of Fisher Discriminant Analysis (FDA) applied to data of blood samples from pregnant mothers who currently have a child with Autism Spectrum Disorder (ASD) and pregnant mothers whose children are typically-developing (TD). Classification returned a 90% accuracy in predicting which group the mothers fall into by using blood metabolite measurements from the Folate-Dependent One Carbon Metabolism and Transsulfuration (FOCM/TS) pathways. The second contribution used logistic regression to find that non-pregnant mothers who have a had a child with ASD and non-pregnant mothers with typically-developing children were able to be classified through the use of general blood metabolite measurements from multiple pathways with 97% accuracy. The third contribution focuses on similar algorithms, but applied to a different problem in that FDA was used on sports-related concussion recovery measurements to find differences between recovered athletes and non-recovered athletes to identify important markers of recovery. This work included balance measurements that have not been widely used before as compared to the standard observational practice that is currently in wide-spread use.
This research highlights the importance of multivariate statistical approaches applied to a few selected life science problems. The approaches are able to find significant measurements in clinical data that may not have been detected through traditional univariate analysis. This work has the potential to change the direction of future research and provide insight into clinical data that may have not been discovered using traditional means.Ph
Accelerated stochastic gradient methods with adaptive techniques and distributed computing
No full textMay 2023School of ScienceStochastic gradient methods (SGMs) have gained widespread usage for solving stochastic optimization problems due to their simplicity and efficient computation as first-order methods. However, vanilla SGMs suffer from slow convergence, prompting the development of many adaptive variants to expedite convergence. With the exponential growth of data, it has become increasingly challenging to process all the data on a single machine within a reasonable amount of time. To address this challenge, leveraging the power of multiple machines in parallel has become an affordable and effective solution to reduce computing time. Despite the popularity of adaptive techniques and parallelization for large-scale data processing, the analysis of adaptive SGMs and distributed methods is often restricted to problems with no constraints or easy-to-project constraints, convex problems, or nonconvex but smooth problems. Many applications are in uncovered forms and remain under-explored, such as Neyman-Pearson classification, fairness-constrained classification, phase retrieval, and sparsity-regularized deep learning. To address these challenges, my research aims to accelerate SGMs by using adaptive techniques in uncharted situations and modify the methods for distributed settings. I have proposed three methods. The first method solves expectation-constrained convex stochastic programs by an accelerated primal-dual SGM. The second method tackles nonconvex (and possibly nonsmooth) programs by an accelerated SGM in a centralized distributed system where derivatives are computed on stale variables. The third method focuses on solving nonconvex stochastic composite problems in decentralized distributed systems with heterogeneous data distributions. These methods have the potential to expand the application of SGMs to a wider range of challenging problems and save the computation time for large-scale datasets.Ph