116964 research outputs found
Sort by
Advancing frontiers of path integral theory for stochastic optimal control
Get PDFStochastic Optimal Control (SOC) problems arise in systems where the dynamics are influenced by random disturbances and uncertainties, such as robotic systems navigating unpredictable environments or financial systems subject to asset price fluctuations. These problems are modeled using stochastic differential equations (SDEs), where the goal is to design control inputs that minimize a cost function while accounting for the probabilistic nature of the system's evolution. Traditional methods like dynamic programming, though powerful, are computationally expensive, particularly for high-dimensional and non-linear systems due to the 'curse of dimensionality'. The path integral control approach offers a promising alternative for solving SOC problems, especially in complex, high-dimensional systems. Based on the Feynman-Kac theorem, the path integral method transforms the SOC problem into an expectation over noisy system trajectories, enabling efficient policy computation through Monte Carlo sampling. As the number of Monte Carlo samples approaches infinity, the policy generated by the path integral method converges to an optimal policy. Due to its purely simulator-driven nature, path integral control is applicable to high-dimensional, nonlinear, and SOC problems, where conventional model-based policy synthesis methods often face challenges. The ability to parallelize the Monte Carlo simulations on Graphics Processing Units (GPUs) makes the path integral approach scalable and well-suited for real-time applications. Additionally, it operates directly on the system’s non-linear dynamics, allowing it to handle a wide range of practical systems where traditional linear approximations fall short. The path integral approach is inherently robust to uncertainties, making it particularly relevant for safety-critical applications such as autonomous vehicle navigation, where accounting for sensor noise and unpredictable conditions is essential for safe operation. Its real-time policy generation capability makes it adaptable to dynamic environments, further enhancing its utility in autonomous systems and robotics. This dissertation develops and applies the path integral control theory to six classes of SOC problems: Chance-Constrained SOC, Two-Player Zero-Sum Stochastic Differential Games, Deceptive Control, Task Hierarchical Control, Risk Mitigation of Stealthy Attack and Discrete-Time Stochastic Linear Quadratic Regulator (LQR). Additionally, it presents a sample complexity analysis of the path integral controller for discrete-time stochastic LQR problems. Through these contributions, this research work lays foundational groundwork for simulator-driven autonomy in solving complex SOC problems, particularly those involving non-linear dynamics and high-dimensional state spaces.Mechanical Engineerin
Essays on selection markets and contract design
Get PDFContract flexibility can improve cost-effectiveness and fairness in labor markets by targeting agents with the largest social benefits. This dissertation extends insights and empirical methods from health insurance markets to three new domains. I demonstrate how without increasing public spending or harming decision-makers, policymakers can target new patients to high-capacity physicians, undertreating physicians to strong incentives, and high-ability students to selective universities.
Understanding the determinants of physicians’ treatment decisions helps inform policies that incentivize appropriate amounts of treatment and mitigate health disparities. In Norway, large administrative datasets and national regulations allow me to measure how primary care physicians heterogeneously respond to shocks and investigate mechanisms. I use this population of physicians for two papers that study shocks to the number of patients and the reimbursement rate. In Chapter 1, I document that physicians spend less time treating existing patients (“crowd-out”) after a quasi-random shock to the total number of patients. I combine heterogeneity analysis with a theoretical model of decision-making to compare explanations and policy responses. Increasing the number of physicians may more effectively reduce crowd-out than incentives for greater treatment per physician. Even with a fixed supply of physicians, targeting assignment of patients to low-crowd-out physicians could eliminate most of the documented effect.
In Chapter 2, I expand on the decision-making model to investigate the social efficiency of contract choice. Typically, physicians all face the same uniform incentives. With a menu of payment contracts, each physician chooses a combination of a piece-rate and fixed fee. Since a regulator does not perfectly observe physicians’ differences or the underlying treatment needs of patients, a menu can help introduce variation in incentives to prevent undertreatment without large increases in public spending or creating perverse incentives for physicians to leave the public system. My empirical framework derives the socially optimal menu and measures its benefits in settings with heterogeneous altruistic agents and unobserved differences in clients’ needs. Generally, the impacts of menu design are theoretically ambiguous; in Norway, my estimates of physician heterogeneity imply that a menu would lead to large welfare gains.
In Chapter 3, I study how more flexible financial aid contracts can improve how students select across universities. Financial aid is often a simple function of income. I estimate the impact of allowing financial aid to vary with a richer set of student characteristics and a future outcome: on-time graduation. I leverage students’ selection on unobserved ability into a historical loan forgiveness program to predict their response to counterfactual financial aid. More flexible contracts lower the price of selective universities for students who are more likely to graduate, increasing statewide graduation. In this context, those students benefit the most from attendance and tend to come from underrepresented demographics.Economic
Iterative Monte Carlo self-shielding factor corrections for neutron activation analysis of rare earth elements
Get PDFThe geographic and geopolitical disparity in the supply and demand of rare earth elements (REE) has triggered urgency in development of new economic sources for REE production. Characterization of samples containing REEs is needed beginning at ore exploration and through all phases of refinement. The similar electron structure between REEs makes them notoriously difficult to chemically isolate. On the other hand, Neutron Activation Analysis (NAA) can characterize material composition with minimal sample preparation. A challenge in accurately assessing REE concentrations using NAA is accounting for neutron self-shielding effects. Isotopes with large neutron interaction cross sections will perturb the neutron flux in and around the sample. Many of the rare earth elements have large neutron cross sections resulting in significant self-shielding.
When the sample constituent mass is unknown a priori, iterative processes are necessary to arrive at an accurate mass and self-shielding estimation. Many analytical and semi-empirical models have been developed to enable fast implementation in iterative loops. The analysis speed achieved in these models comes with the expense of simplifying assumptions and constraints. The Monte Carlo approach is capable of improving many common error introducing simplifications including geometry constraints, resonance modeling and interference, scattering modeling, gamma attenuation, empirical correction factor dependence, and curve fit approximations. However, Monte Carlo models are often dismissed as a viable solution for routine NAA analysis due to computational expense.
A tailored MCNP model has been developed and implemented in an iterative loop to estimate the constituent mass of unknown experimental samples. In challenging applications, with neutron self-shielding near 50%, the tailored neutron and gamma self-shielding models were able to achieve convergence in 9 iterations and 13 minutes of MCNP computation or less. In applications with minimal self-shielding convergence can be achieved in under 2 minutes.Mechanical Engineerin
The impact of permeability and elastic properties on seismicity in the central Hikurangi Margin
Get PDFThe physical properties of subducted sediments impact the deformation of subduction zones and can control the occurrence of slow slip events (SSEs). The northern part of the Hikurangi Margin exhibits unusually shallow slow-slip events while central and southern section’s appear locked within the shallow depth range (0-25 km). In particular, SSEs along The Hikurangi Margin occur at different time scales and depths when comparing the northern and southern ends of the margin. Previous studies show that the rock comprising the accretionary wedge of the northern margin has low permeabilities (<100 µD), which could control the over pressurization of under-thrust fluids, modulate the occurrence of SSEs and are an important factor for wedge evolution. Permeability rises when an SSE fractures the rocks within the deep wedge promoting fluid flow and thus dissipating the overpressures along and above the décollement. As fractures heal and permeability recovers overpressures build up once again. Although this cycle may explain the occurrence of SSEs along northern Hikurangi, it is not yet clear how intrinsic permeability varies in rocks above the décollement elsewhere along the margin.
Rock samples from the northern and central part of the margin have been tested for permeability and elastic properties to better understand the disparity in SSE occurrence. We tested samples from the Weber (PQ), Whangai (PO), Wanstead (WP) formations and the Waioeka facies (DV), which are representative of the lithologies above the décollement in the central margin, and range in age from the Cretaceous to the mid to late Paleogene. We used a pressure vessel equipped with a pore pressure fluid circuit to measure permeability and ultrasonic velocities over a range of confining pressures (20-200 MPa). We found that the Weber formation (PQ) had a permeability of around 0.3 µD and Whangai formation (PO) had a permeability of around 0.6 mD, both of which are from the central Hikurangi Margin and clearly show higher permeabilities than the northern Hikurangi Margin rocks from the same geologic age. Higher permeabilities for central HM samples promote the idea that the central Hikurangi Margin accretionary wedge can efficiently drain fluids that are subducted rather than confine them. Suggestions as to why this higher permeability is present range from differing times in which the subduction initiated and differing grain sizes from a contrast in sediment source. This study provides insight into the mechanisms that lead to significantly different slip behaviors along the central Hikurangi Margin and can be utilized in a wide range of other study areas.Earth and Planetary Science
Borges, Jung, and Memory: The Circular Ruins of Cognition
Get PDFMuch has been written about the short stories of Jorge Luis Borges and Carl Gustav Jung in isolation. Furthermore, scholars have analyzed the role of memory in Borgesian fiction and Jungian psychology. This thesis aims to synthesize these three things (Borges, Jung, and memory) and identify the interplay between them. I will argue that the works of Jorge Luis Borges complement the Jungian school of thought and the phenomenological tradition, placing an emphasis on memory’s role of unifying the mind and the body. I will identify specific areas of overlap between the works of Borges and the thought of Jung using memory as a framework. The two figures agree on three important elements of memory: its intentionality, its morality, and its potential dangers.Spanish and Portugues
Practicing the Labyrinth: A Social Analysis of Constrained Literature
Get PDFThis thesis offers a social and historical analysis of constrained literature, where constrained literature is defined as literature subject to any set of explicit rules imposed on its production. The purpose of a historical analysis is to explore continuity and change in the social implications of literary constraint over time, with similar constraints such as the sonnet having different connotations over the course of their histories. This thesis mainly focuses on the Oulipo, a literary group founded in 1960 with the explicit goal of navigating the potential of literature through rigorous constraints, but I look at contemporary examples of literary constraint as well. The Oulipo do not have a central literary or political ideology, and this thesis discusses whether or not there are inherent social affordances of constraint and particularly the way in which the Oulipo practice writing under constraint. I apply Pierre Bourdieu’s A Theory of Practice and Distinction as my primary sociological framework in interrogating the symbolism of constraint and the social contexts related to form and representation in literature. Bourdieu’s theory reveals contradictions in the notion of constrained literature as a means of democratizing the writing process and as an equalizing force. However, I connect the Bourdieusian and Oulipian notions of practice in order to show how Oulipian constraint, in focusing on the act of writing rather than the theory of writing, is particularly well suited to dealing with these contradictions.Plan II Honors Progra
Ballad of an immigrant with memory
Get PDFThis report details the process of creating Ballad of an Immigrant with Memory, a narrative short film produced as part of my MFA thesis in Film and Media Production. The film follows an undocumented Mexican immigrant in the U.S. who, after submerging into a bowl of menudo, embarks on a surreal journey through his memories of Mexico in search of the rosary his mother gave him before he left the country 30 years ago. In the following pages, I reflect on the film’s development, production, and post-production processes. I also explore how my team and I approached the film’s magical realism elements—particularly through art direction, cinematography, and visual effects. Finally, I offer insight into my personal creative process and the post-production outcomes that shaped the final version of the project.Radio-Television-Fil
Invisible margins : speculative wearables challenging social norms for introverts
Get PDFIn contemporary society, social norms are designed around the extrovert ideal, where social engagement is expected, and withdrawal is often perceived as avoidance or failure. This structure creates significant challenges for introverts, who frequently experience social anxiety, conformity pressure, and identity struggles. While mainstream social environments prefer sociability and ease in public settings, they often fail to accommodate diverse interaction styles.
Invisible Margins is a speculative design project that challenges these dominant norms by proposing an alternative reality, one where social interaction is not dictated by strict expectations but instead adapts to individual comfort levels. Through three wearable designs, this project explores emotional connection, personal space, and gaze perception, questioning the assumptions embedded in mainstream social behaviors.
Hugget is an inflatable wearable that provides the feeling of hugging without human touch. It explores how emotional support can take alternative forms when conventional physical touch is not viable. Refudress is a wearable that can inflate into a personal retreat in public space as the wearer withdraws into it. It challenges the stigma attached to stepping away by reframing it as a strategy of self-regulation and resilience. Obscura is a pair of perceptual glasses that mediate eye contact and visual exposure with interchangeable lenses. It explores how self-regulated visibility might shift the power dynamics embedded in the act of looking.
This project shows the psychological reality of introversion and questions the supremacy of extrovert-centric paradigms of social interaction. Rather than offering prescriptive solutions, it envisions new social imaginaries that validate withdrawal, reframe silence as a form of agency, and acknowledge introversion as a vital dimension of human diversity.Desig
Numerical investigation of bond deterioration in ASR-affected reinforced concrete beams with lap splices
Get PDFAlkali-Silica Reaction (ASR) is a deleterious concrete deterioration mechanism whereby alkali hydroxides react with certain aggregates, resulting in the formation of an expansive gel that induces micro- and macro-cracking within the concrete. A critical consequence of ASR deterioration is the loss of bond integrity between concrete and reinforcing steel. While existing bond models in the literature were found to accurately predict behavior of sound concrete specimens, not affected by ASR, they failed to capture the effects of ASR-driven bond degradation—especially in regions with lap splices, where structural continuity depends entirely on effective bond.
This thesis addresses the need for improved characterization of bond behavior in ASR-affected reinforced concrete beams with spliced bars. First, existing bond- slip models are examined, highlighting their range of applications and limitations when applied to deteriorated concrete. In addition, a comprehensive review of experimental studies is presented, establishing a benchmark for validation. The core of this research involves a comprehensive analytical study consisting of nonlinear finite element (FE) analyses of sound concrete and ASR-affected beams with spliced bars.
A novel bond stress-slip model is proposed to account for ASR-induced degradation. The model is implemented within VecTor2 using a three-point reference approach, which allows tailored bond degradation characteristics. Validation studies for ASR-affected concrete were performed through comparisons of calculated and observed load-displacement responses for both reactive and non-reactive beams. Results show that, for spliced reinforcement under ASR, the new bond model significantly improves correlation with experimental data, reducing errors in peak load prediction and capturing bond failure patterns more accurately.
This thesis provides a validated framework for more realistic analysis of ASR-affected RC members with lap splices. It underscores the importance of accurately modeling bond deterioration in aging infrastructure. Future work will extend these findings by refining the descending branch of the bond-slip relationship and exploring long-term ASR progression in large-scale structural simulations.Civil, Architectural, and Environmental Engineerin
Scaling laboratory ASR expansion measurements for application in structural concrete modeling
No full textThe research work presented within this thesis is meant to provide a rational method to estimate the expansion for any given concrete structure that is affected by Alkali-Silica Reaction (ASR). The research program attempts to use free expansion data from standard ASTM C1293 prisms to evaluate the expansion in orthogonally reinforced cube specimens, cast with the same concrete mixture. It is known that the ASR expansion is volumetric in nature, and it is influenced by the presence of reinforcement and boundary conditions, alongside the geometrical differences of concrete specimens. An adjustment is made to the collected prism data with reference to the gauge lengths used for comparison. Using the adjusted free expansion data, an analysis was conducted for specimens including panels and beams to test the validity of the proposed method for evaluating realistic levels of expansion, as well as the impact on the calculated failure modes. While the method is not presented as a definitive solution for modeling ASR expansion, it serves as a foundational step toward the development of more robust, accurate tools for assessing both existing and future structures impacted by ASR.Civil, Architectural, and Environmental Engineerin