137091 research outputs found
Sort by
Impact of Content Domain and Preference on Pedagogical Agent Effectiveness
No full textPedagogical agents (PAs) are virtual, typically anthropomorphic, characters used in online learning environments to improve users’ learning outcomes. They have been shown to increase learning and motivation and are also suggested to reduce cognitive load. Despite the widespread use and investigation of PAs, questions remain. One is: does a learner who prefers not to learn with a PA reap the same pedagogical benefits as one who does? To explore this, the effects of matching versus mismatching PA usage with learners' preferences were analyzed. Additionally, whether PA effectiveness varies across content domains (examined through mathematics and art) was investigated. This study comprises two experiments. Experiment 1 used a 2 (PA Assignment: With-PA vs. No-PA) by 2 (Lesson: Math vs. Art) mixed design to assess immediate learning, retention, motivation, subjective workload, PA persona ratings, and opinions. Experiment 1 was designed to (a) evaluate the PA’s effectiveness prior to Experiment 2, and (b) compare its impact in a math lesson versus an art lesson. Experiment 2 used a 2 (PA Preference: Prefers vs. Does Not Prefer) by 2 (PA Assignment: With-PA vs. No-PA) between-subjects design. The same dependent variables (excluding retention) were used to analyze the effects of preference-assignment congruence. No hypotheses were supported, and, notably, the main effect of PA Assignment was nonsignificant across both experiments. Significant main effects of Lesson (Experiment 1) and PA Preference (Experiment 2) were found. Implications, limitations, and avenues for future research are discussed
Vertically integrated end to end technology evaluation platform for CMOS and beyond CMOS
No full textThis work presents a comprehensive end-to-end evaluation platform for CMOS and beyond-CMOS technologies based on experimentally validated/calibrated models. This work encompasses various levels of abstractions while considering various device, interconnect, and memory options across multiple technologies. In the case of memories, the focus has been on spintronic memories including spin orbit torque (SOT), spin transfer torque (STT), and magnetoelectric (ME) based magnetic random access memories. For SOT devices, a comprehensive modeling for spin current generation in nanoscale devices is presented while accounting for the impact of non-uniformity in the electric current density and material properties. For SOT and STT devices rare-event enhancement based methodology has been adopted to evaluate various trade-offs among write error rate (WER), delay, and current. Area saving schemes for SOT-MRAM, with multiple magnetic tunnel junctions (MTJs) on a SOT track using voltage-controlled magnetic anisotropy (VCMA) and STT, are presented while accounting for the detailed trade-offs among voltage drop, WER, energy, and bit density. A study on the impact of scaling technology node from the 14nm node to 7nm node is presented for MRAM arrays while also evaluating the impact of various back-end-of-line (BEOL) technology options. In addition, at the 7nm node full SOT-MRAM memory system design is presented based on open-source ASAP7 process design kit (PDK). To explore the applications SOT/ME devices for hardware accelerators, novel content addressable memory (CAM) designs are proposed and evaluated. At the 3nm technology node, a PDK has been developed to evaluate various device, interconnect, and technology options. Using gate-all-around field effect transistor (GAAFET) based standard cell library, various BEOL options involving copper and ruthenium are studied using place and route results for benchmark circuits.Ph.D.Electrical and Computer Engineerin
Microbubble Dynamics Monitoring and Control for Diagnosis and Treatment of Brain Cancer
No full textMicrobubble-enhanced focused ultrasound (MB-FUS) is an emerging technology
especially for drug delivery in the brain as it enables non-invasive, reversible, and targeted physical
opening of the blood-brain barrier (BBB). MB dynamics (i.e., radius change as a function of time)
during FUS excitation plays a key role in the efficacy and safety of this procedure. Thus,
reproducible tuning and monitoring of the MB dynamics to an effective strength while avoiding
collapse of MB, which has been associated with tissue damage, is essential. Several past studies
have proposed algorithms to control MB dynamics in the brain. However, a robust method that
ensures safety and adapts to the highly non-linear bubble oscillation and dynamic environment in
the brain remains an open problem.
The overall objective of the thesis is to design and evaluate control methods to attain
desired MB dynamics in the brain using their acoustic emissions (e.g., spectral contents). Specific
aims of the project are 1) to design an acoustic emission-based algorithm to control MB dynamics
in the brain and assess the performance and efficacy in healthy brains and glioblastoma (GBM)
tumor model (GL261) in rodents; 2) to improve the safety of the controller using machine learning
methods, and 3) develop acoustic based methods to directly monitor MBs’ radius in in vitro
conditions.
By bridging the MB dynamics with their biological responses, this work will not only refine
our understanding on the role of MB dynamics during MB-FUS technology, but also support the
development of novel diagnostic (e.g., liquid biopsy) and therapeutic strategies against brain
diseases (e.g., brain cancer immunotherapy). Furthermore, in combining MB-FUS treatments with
existing therapeutic strategies, the ability to control the MB dynamics using acoustic methods will
ultimately support their safe application in humans along with the widespread adoption of this
technology.Ph.D.Mechanical Engineerin
Computational Electrosynthesis: The Role of the Electrical Double Layer in Anodic Organic Transformations
No full textElectrosynthesis is the use of applied potentials in an electrochemical cell to drive electron transfer and form new chemical compounds. These transformations can proceed through oxidative or reductive pathways, typically through radical or radical-ion intermediates. Because these intermediates are high in energy and involve relatively low reaction barriers, they are very effective at enabling transformations that are difficult or inaccessible under purely thermal conditions. At the same time, this reactivity can make electrosynthetic reactions difficult to control and has historically contributed to low yields and poor selectivity. As a result, electrosynthesis has remained relatively underutilized in organic synthesis, especially compared to what its synthetic potential would suggest. Despite these challenges, electrosynthesis also offers many benefits. The use of electrical current as a means to drive redox reactions makes it a direct application of renewable energy initiatives, and it is considered more environmentally friendly or "green" as compared to traditional oxidative/reductive chemistries which make use of stoichiometric chemical reagents. Through the use of flow cell geometries, reactions developed at the bench can be quickly scaled up to meet industrial applications. For these reasons, electrosynthesis is currently experiencing a "renaissance," with many research groups revisiting classic electrosynthesis reactions and developing innovative new methods for improving their selectivity and yield. Of particular interest is the understanding how environmental parameters, such as choice of solvent, supporting electrolyte, electrode material, and overpotential, can be used as knobs to control the underlying mechanisms of these kinetically controlled reactions. The lifetimes of these radical-ion intermediates are typically very short, on the order of nanoseconds, making them difficult to probe directly via experiment. For this reason, quantum chemistry, statistical mechanics, and molecular simulation approaches present a promising avenue for exploring how the chemical reactivity, diffusion properties, and redox potentials of reactive intermediates depend on the electrochemical environment. The focus of this thesis is on the use of molecular dynamics and density function theory-based quantum mechanics/molecular mechanics simulations coupled with free energy sampling methods to understand how the electrode-electrolyte interface modulates the reactivity of radical-ion species, with a particular focus in electro-organic reactions proceeding via anodic oxidation
Towards Born Qualification of AM Components: High Temperature Fatigue Testing and Microstructural Characterization of AM IN718
No full textThe high temperature mechanical fatigue testing of additively manufactured (AM) IN718 specimens created using varying process parameters was conducted. The microstructure of the AM specimens was characterized using x-ray computed tomography and electron backscatter diffraction (EBSD). Multiple machine learning predictive algorithms were implemented on the assembled training data compiled from testing and characterization. Results showed a negative correlation between the total amount of porosity within the gage region of the specimens and their corresponding fatigue lives. Though, for specimens with lesser amounts of porosity, the presence of high criticality porosity, defined by pore size, shape, and proximity to other pores and the specimens surface, directly correlated to diminished fatigue performance. Results of the testing and characterization also showed the fatigue properties are defined by the competitive influence of crystallographic and porous microstructural features. The competitive influence of the crystallographic microstructure on the fatigue properties was well captured by a EBSD derived parameter termed the areal average resolved shear stress. Machine learning methods, which were examined using Shapley value feature importance, reflected the same competitive influence between crystallographic and porous features in the accurate prediction of fatigue properties. These findings lay the groundwork for defining the process-structure-property relationships necessary to enable born qualification of AM IN718 components for high temperature applications.Ph.D.Materials Science and Engineerin
Motor unit control of mouse locomotion
No full textLocomotion relies on neuromuscular circuits that produce rhythmic activity across the limbs and body. Motor neurons within the spinal cord innervate limb muscles, causing them to contract and generate the forces that moves the body. For movements as complex as locomotion, motor units, which consist of a single motor neuron and the muscle fibers it innervates, must be coordinated within and across muscles to not only move, but also to flexibly adjust locomotor rhythm. Although it is well established that motor units modulate the force output in muscles largely through their recruitment and firing rate, it is unclear how these firing patterns, both in individual motor units and in populations of motor units, are coordinated during dynamic movements. Using novel neurotechnology that allows for high-resolution investigation of the neuromuscular system, work presented in this thesis describes the coordination of motor units during locomotion in mice across different speeds. In Chapters 1 and 2, we introduce key concepts regarding the neuromuscular control of locomotion and review literature in the field. In Chapter 3, we identify how the firing rate and recruitment probability of individual motor units correlates with not only movement speed, but also specific kinematic features of each stride. Characterizing these results across different muscles in the mouse forelimb, we highlight how muscles with shared functions may be uniquely controlled by the nervous system. In Chapter 4, we use pair-wise analyses of motor units to determine how populations of motor units are recruited and de-recruited within muscles. Motor units were recruited but not de-recruited in a systematic order, and deviations from this order were correlated with stride-by-stride changes in locomotor behavior. Taken together, these results provide evidence on how the mouse neuromuscular system coordinates robust and flexible locomotor behavior.Ph.D.Biomedical Engineerin
A Model-Based Approach to Understanding and Improving Video Conferencing Quality of Experience
No full textIn recent years, influenced by trends towards remote work, video conferencing has be-
come an important tool for business, education, and personal communications. Network
operators, business users, and researchers have a renewed interest in how video conferenc-
ing applications (VCAs) work, and improving the user quality of experience (QoE).
We look at the question of VCAs and QoE from two perspectives; networking and
interactivity. On the networking side, we build an understanding of VCA architecture by
developing functional models of the VCA client and selective forwarding unit (SFU), the
central server that forwards media between users and enables efficient scaling of video
conferences. We developed these models using an active measurement campaign to study
aspects of the VCA and SFU such as the congestion control and user interface interaction.
On the interactivity side, we designed a model of how a VCA client takes turns and
interacts with others. Combined with objective metrics we define for interactivity, we
can experiment with network latency and understand the negative effect it has on these
metrics and hence the QoE. This model-based methodology expands upon the user study
method commonly employed by literature in this area. By implementing appropriately pa-
rameterized client models within a simulator program, we show how latency impacts the
interactivity metrics, such as the overlap rate or the proportion of useful conversation time.
Lastly, we investigate how to improve interactivity in the presence of network latency.
We use the client model methodology to evaluate systems suitable for deployment on the
VCA SFU that can address the issue of network latency by improving the interactivity
metrics. In particular, we design and evaluate two systems; one that adjusts the latency of
the forwarded media, and one that informs clients of latency via notifications, and show that
these methods can improve interactivity across a variety of video conferencing scenarios.
We hope that the models and results from this research will be useful for improving
user QoE for video conferencing, an important and modern communications tool
Formulation and Process Considerations for the Manufacturing of Dense Pastes
No full textAchieving high-quality functional composites with reproducible properties is a challenge as dense pastes are highly sensitive to formulation factors and processing conditions. These challenges can be mitigated through the implementation of quality by design (QbD) principles, including the development of formulation–process–property relationships and process analytical technologies (PATs). In this dissertation, challenges associated with dense paste mixing and shaping are addressed through the development of PATs and quantitative formulation–process–property relationships that enhance process understanding and material knowledge. Central to this work is the use of Resonant Acoustic® Mixers that employ resonant vibrations and are well suited for mixing highly viscous dense pastes. I demonstrate that motor data can be leveraged as an in-situ PAT capable of detecting changes in mixing behavior arising from both formulation and processing effects, including variations in total solids content and lot to lot material variability. I further show that transitions between coupled mixing and splashing behavior depend directly on paste viscosity and adhesion to the mixing vessel. I characterize these dependencies using a novel dimensionless adhesion number, AD, which captures the balance between kinetic input, viscous dissipation, and adhesive interactions. I also evaluate mixer diagnostics from a simplified model system of rubber bouncy balls to validate these measurements and establish clear guidelines for data interpretation and post-processing. Finally, I investigate the effects of polymer molar mass in the binder on heterogeneity formation during processing. I show that low molar mass binders at high concentrations provide improved stability against settling and shear-induced migration, which is attributed to balanced viscous dissipation and elasticity. Collectively, the findings presented in this dissertation establish a framework for improving process understanding and material knowledge to overcome critical challenges in the manufacturing of dense paste composites