12023 research outputs found
Sort by
Machine-Learned Propulsion Strategies: From Adaptive Damage Compensation to Advanced Aeromobility
No full textAutonomous vehicles are regularly sent into "dull, dirty, and dangerous" environments where the risk of damage is high. Avoidance or mitigation of such damage is therefore paramount to maintain effective autonomy. In this thesis, we use machine learning to investigate two different propulsive strategies that may be used by autonomous vehicles. The first, flapping propulsion, shows remarkable ability in nature to recover from damage simply by altering stroke kinematics. Using machine learning, we ask whether and how such mitigation of damage would be possible for a robotic autonomous vehicle. The second propulsive strategy we investigate is single-rotor propulsion, most commonly seen in helicopters. With this system, we seek to avoid damage before it occurs by improving mobility and control authority via thrust vectoring.
In Part I, we use an evolutionary strategy (CMA-ES) with hardware-in-the-loop to explore optimal machine-learned adaptations to propulsor damage. Experimental function evaluations are performed by a flexible propulsor actuated by a spherical parallel manipulator (SPM). The machine-learned forces and trajectory parameters are compared to in vivo observations in order to determine whether bio-inspired strategies to adapt to significant propulsor damage are the most efficient, or whether they may be affected by irrelevant evolutionary pressures. With amputation of approximately 50% of the propulsor, we find that a complete recovery in thrust production and fitness is made. Some characteristics of the recovered trajectory are similar to natural swimmers, while others differ. Recovery when producing side-force is even more complex. Not all trials are able to recover force production and fitness, and no clear strategy to modify amplitude or frequency is seen. We conclude Part I by using PIV measurements to detail the effect of compensatory strategies on hydrodynamics. Both amputated and intact trajectories clearly show utilization of a drag-based paddling strategy, but the hydrodynamics of the intact and amputated fins differ significantly. This suggests that the machine-learned trajectories are not simply reestablishing the same wake as the intact fin to achieve the same thrust and fitness.
Given the success in applying machine learning in-the-loop to a complex propulsive system where fluid-structure interactions are significant, we utilize the same strategy in Part II to begin to explore helicopter aeromechanics. We built an independent blade control (IBC) system that interfaces with the CMA-ES algorithm to explore optimal blade pitch trajectories. Using this platform, we explore two preliminary optimizations designed to vector thrust; the first, for sustained thrust vectoring that might be utilized upon takeoff or landing, and the second, for short-time thrust vectoring that could be used for enhanced maneuverability. We present some preliminary results from these optimizations and lay out a foundation for future applications of this experimental system.</p
Global Analysis of Protein Synthesis and Degradation in Escherichia coli
No full textProtein synthesis and degradation shape the cellular proteome to drive homeostasis and physiological adaptation. Many fundamental aspects of protein regulation have been elucidated through investigation of the Gram-negative bacterium Escherichia coli, which remains a fruitful model organism for uncovering conserved regulatory mechanisms relevant to cell biology, biotechnology, and medicine. Here, we used bioorthogonal noncanonical amino acid tagging (BONCAT) for the time-resolved analysis of protein synthesis and degradation in this organism in several contexts. We profiled protein degradation on a proteome-wide scale in growing and growth-arrested cells, identifying instability in a diverse panel of regulators. Our identifications served as training data in the validation and deployment of a machine learning classifier of in vivo protein stability, which highlighted the role of active degradation in motility and surface adhesion. We then utilized an efficient system of active degradation in this organism to engineer the instability of the mutant methionyl-tRNA synthetase NLL-MetRS for the analysis of protein synthesis in transient physiological states. Destabilized NLL-MetRS variants exhibited half-lives on the order of hours, which improved the fidelity of metabolic labeling in growth-arrested cells. Additionally, we leveraged the sensitivity of BONCAT to investigate protein synthesis in growth-arrested cells expressing a well-studied but controversial member of the widespread toxin-antitoxin family, MazF. Our proteomic profiling suggests this toxin activates several endogenous stress response systems, most notably the cold shock response system. Taken together, these investigations highlight the advantage of time-resolved proteomics in characterizing proteome dynamics
Part I: Multi-Valent Ion Effects on Polyelectrolyte Structure and Thermodynamics & Part II: Hydrodynamic Self-Propulsion
Get PDFPart I: Polyelectrolytes are a class of charged polymers that have found widespread utility in water treatment, drug delivery, and scale inhibition, among other applications. For many of these applications, it is crucial to control the phase stability of the polyelectrolyte solution. The long-ranged nature of the electrostatic interactions in polyelectrolyte solutions and the polyelectrolyte's connectivity lead to a rich phase behavior that can be challenging to study, especially in the presence of other ions or surfaces. In scale inhibition applications, polyelectrolytes such as poly(acrylic acid) (PAA) are used to prevent the dissolution of sparingly soluble salts, such as calcium carbonate, in water. While the significant influence of small ions on polyelectrolyte solution phase behavior is recognized, the precise molecular mechanisms driving the resulting phase stability remain largely elusive.
Polyelectrolyte theory suggests that a polyelectrolyte's behavior and adsorption properties in solution are strongly tied to the polymer chain conformation and charge distribution, which in turn is influenced by solution ionic strength and ionic valency. Consequently, we expect the polyelectrolyte performance to be highly dependent on the solution conditions and the molecular features of the polyelectrolyte. Previous computational studies have studied general polyelectrolytes in solution with coarse-grained and implicit solvent models and provided insights into the chain conformational transitions. However, they disagree on the mechanisms underlying aqueous polyelectrolytes salting out of suspension and are unable to yield chemically specific insights. We seek to better understand the antiscalant mechanisms of polyelectrolytes using all-atom molecular dynamics to capture solvation and polymer chemistry effects on the mechanisms of polyelectrolytes preventing scale nucleation and slowing growth.
The current work investigates the structure and thermodynamics of polyelectrolytes in bulk solution and at crystalline interfaces with added multi-valent ions. The presence of multi-valent ions, such as Ca2+, can significantly influence polyelectrolyte conformation via ion bridging non-neighboring charged monomers as well as screening the electrostatic interactions. We employ all-atom molecular dynamics simulations to investigate the binding modes of Ca2+ onto a PAA chain, Ca2+–PAA complex aqueous stability, and PAA adsorption onto a crystalline CaCO3 surface. In each of these cases, we find that the balance between ion bridging, electrostatic screening, and water-mediated interactions plays a crucial role in determining the polyelectrolyte's behavior in solution and at an interface.
Part II: Active bodies undergo self-propulsive motion in a fluid medium and span a broad range of length and time scales. Many active systems spontaneously self-organize into visually striking structures: fish schooling, birds flocking, and bacterial colonies growing. Current models of this emergent behavior in the inertial regime are mainly phenomenological and lack consideration of the fluid-mediated interactions between bodies.
To address this limitation, we seek to advance physics-based models of swimmers by explicitly incorporating the fluid mechanical interactions between bodies. We aim to discern the fluid medium's role in group dynamics and determine whether it can reproduce the observed emergent phenomenon without resorting to phenomenologically based interaction rules. To that end, we focus specifically on swimming in high Reynolds number flows, where inertial forces dominate, and draw comparisons to the well-studied low Reynolds number (Stokes) regime. We begin by deriving the equations of motion for a collection of unconstrained spherical particles in potential flow and extend the model to include viscous dissipation and rigid body motion constraints for many bodies with arbitrary kinematics.
We then consider the case of a single swimmer consisting of three linked spheres in potential flow. Through this, we find self-propulsion without needing external forces or momentum transfer via vortex shedding. We compare the inertial swimmer to an identical swimmer in the Stokes regime—where fluid inertia is neglected—and find that the structure of the equations of motion is identical in both flow regimes. Notably, the Stokes hydrodynamics are longer-ranged at leading order, leading to a more significant net displacement of the swimmer after one period of articulation. Finally, our study provides analytical insight into the swimming of a deformable body through an expansion of the non-linear spatial dependence of the hydrodynamic interactions.</p
Modeling Frictional Processes in the Presence of Fluids: From Earthquakes in the Laboratory to Induced Seismicity in Geothermal Reservoirs
Get PDFInduced seismicity - earthquakes driven by injections of fluids into the subsurface - is of growing societal importance in its impact on clean energy technology. Advancements central to the world’s transition to a greener economy such as geothermal energy and long-term geologic storage of CO2 are hampered by a lack of understanding and control of the associated seismic hazards. In its mechanics, frictional processes in the presence of fluids is a difficult problem to model given the challenges of studying frictionally unstable material in a controlled environment. Unstable gouge material is commonly found along faults in nature, due to pulverization of brittle rock in to granular layers called `gouge.' This thesis approaches the challenge at two different scales: 1. at the scale of the localized shear layer along the interface between two faults where we model laboratory earthquakes in the presence of pressurized fluids, and 2. at the scale of a reservoir where we model the rate of earthquakes given the injection/extraction schedule.
In order to infer the frictional properties of unstable gouge material from laboratory experiments, we develop a probabilistic model based on a spring-slider representation of the experiment along with the rate-and-state friction law. Inversions indicate that the presence of pressurized pore fluids stabilizes the gouge - by an increase in the strength of the contacts and a lesser decrease in the grain size with slip - even under the same effective normal stress. Assuming purely slip-dependent healing of friction leads to an evolution of parameters with slip that is consistent with previously established interpretations of rate-and-state parameters. The best fitting spring-slider model still shows significant discrepancies to the experiment in the evolution of creep and in the dependence on loading rate. A quasi-static finite-element model with the same rate-and-state properties suggests that the gouge in the sample likely slides in a spatially uniform manner. Thus, the discrepancies between the spring-slider model and the experiment can likely be attributed to flaws in the rate-and-state formalism and the slip law rather than the idealization of a finite geometry to a single-degree-of-freedom system. The results prove that quantitative analysis of frictional processes of gouge in the unstable regime is possible, and that future development of constitutive relationships for friction should aim to reproduce key features of stick-slip in detail.
To model seismicity induced by a geothermal well stimulation, we develop physical and statistical models of the seismicity rate. The physical models are based on rate-and-state friction and stress changes due to pore-pressure diffusion. The statistical model performs a convolution of a kernel function inspired by Omori law decay with the injection rate. Both models successfully reproduce the seismicity observed during the 2018 enhanced geothermal system (EGS) simulation in Otaniemi, Finland. We find that the effect of time-dependent nucleation from rate-and-state friction is crucial in reproducing the temporal and spatial patterns of the observed seismicity. We also find that the effect of finite nucleation cannot be approximated well by introducing a stress threshold in the standard Coulomb friction model, at least in the context of rapid variations of injection rates common in EGS operations.
We highlight the major assumptions of the Dieterich seismicity rate model and examine how they may bias interpretations of induced seismicity observed in real reservoirs by comparing it directly to a Discrete Fault Network (DFN) model. The spatio-temporal pattern of seismicity in the finite setting is not only dependent on fluid transport properties and its combination with nucleation characteristics but also the distribution of initial conditions of the fault network. The back-propagation front, in particular, occurs co-injection if the time to instability for the minimum slip rate is shorter than the injection duration. The relocated catalogue of the 1993 GPK1 stimulation in Soultz-Sous-Forets shows such a back-front which can be fit qualitatively using the time to instability measure. A simple model for the rate of magnitudes that accounts for the evolution of frictional stability reproduces the apparent increase in the source radius of induced events in Soultz-Sous-Forets. The rate of larger events is overestimated by the model, possibly due to an overestimation of maximum magnitudes by the volume of stimulation. The comparisons reveal that parameters of the Dieterich model lack clear physical meaning in the finite analogue and highlight the importance of using realistic physics, especially in models at large scales where uncertainty due to assumptions at smaller scales may be amplified.
We end the thesis with the application of rate-and-state friction to dynamic rupture modeling of seismic data from distributed acoustic sensing (DAS). The modeling of the high-frequency DAS recordings of a Magnitude 6.0 earthquake suggests a highly heterogeneous underlying fault with several prominent asperities and barriers that may control rupture dynamics. The model demonstrates how the high-stress patches both inhibit and promote the overall rupture, while also contributing to a significant amount of the energy release themselves. The successful interpretations of modern seismological data encourage future development efficient models that can be used for dynamic inversions.</p
Engineering Bioaffinity Sensors toward Continuous Electrochemical Biosensing
No full textThe rise of wearable sensing through smartwatches and continuous glucose monitors has made health data more widely accessible. Advances in machine learning have also been pivotal in identifying personalized health insights from biometric data streams. However, continuous biochemical data has been limited in sensor design by the availability of oxidoreductases (e.g., glucose oxidase, lactate dehydrogenase) to a given target. The challenge in engineering diverse oxidoreductase enzymes has led to the exploration of other generalized approaches to continuous electrochemical biosensing. To meet this need, we have explored a variety of bioaffinity sensing schemes using broad bioreceptor classes including antibodies, nucleic acids, and periplasmic binding proteins. We present a case study in electrochemical sensor design utilizing high-affinity antibodies for the rapid diagnosis of COVID-19 disease states. We then investigate the potential of nucleic acid-based electrochemical sensors for continuous sensing with a focus on structure-switching nucleic acid aptamers. The utility of aptamer sensors is demonstrated in the development of a serotonin aptamer sensor embedded in an ingestible capsule for continuous biosensing in the gastrointestinal tract. Applying the principles of electrochemical aptamer-based sensing, we explored the development of an electrochemical protein-based sensor for nicotine, which exploits the hinge-like binding motion of periplasmic binding proteins while also capitalizing on decades of protein evolution and characterization research. With the goal of continuous, noninvasive biochemical sensing, we evaluate the design considerations and translatability of these sensors for wearable sweat analysis. These biosensing techniques may enable the future hardware necessary to expand accessible biomedical data for the next wave of personalized health monitoring
Planning for an Uncertain Future: Tree-Based Methods for Real-Time Fault Estimation, Collision Avoidance, and Multi-Agent Reconfiguration
Get PDFAutonomous spacecraft making independent high-level decisions present the promise of dramatically increased productivity in space for both exploration and economic activity. While autonomy has seen limited use in space to date owing to a lack of flight heritage, limited computational resources, and a traditionally risk adverse industry, the growing numbers of spacecraft and increasingly ambitious missions will soon render the current ground-intensive mode of space operation untenable.
In this thesis, we develop two critical capabilities for an autonomous future in space. The first is proactive fault estimation, which seeks to rapidly and safely identify the root causes of onboard anomalies by planning sequences of test actions to gather information while probabilistically ensuring safety. The second is real-time reconfiguration to enable formations of spacecraft to respond quickly and effectively to changing environments or mission objectives.
We achieve both goals using various forms of Monte-Carlo Tree Search planning. By formalizing each capability as sequential decision-making problems, and developing algorithms well suited to information gathering, we show that our algorithms provably converge to optimal solutions while maintaining the ability to run in real-time on robotic spacecraft simulators. We present several algorithmic innovations, including marginalized filtering, sampling-based chance constraint evaluation, and an array-based implementation of Monte-Carlo Tree Search. Through and numerical simulations and hardware experiments, we demonstrate that these modifications enable our algorithms to outperform existing tree search methods and achieve better scaling across system complexity, noise, and simulation depth.</p
From Daily Deformation to Millennial Mechanics: Insights from Subduction Zone Earthquake Cycle Models
Get PDFSubduction zones have hosted all the largest five earthquakes in the last one hundred years, including the 2011 Mw 9.1 Tohoku-oki earthquake on 11 March 2011, one of the largest natural disasters in history. While the general mechanism of these thrust-style earthquakes is well described by stress accumulation due to the locking between the incoming and the overriding tectonic plates, many questions remain as to the size and longevity of the asperities which host the coseismic rupture, the rheological models best describing the rock in and around the fault zone, and the effects of stress shadows and interactions between different asperities on the same plate interface. These questions are addressed by using large earthquakes as natural experiments, which we can observe using geodetic, seismic, and other techniques. However, ambiguities in the modeling results point to the inherent problem of non-uniqueness when interpreting surface observations of single events to infer complex processes at depth.
This dissertation presents a new framework to study subduction zones and their rheological properties by extending both the time period modeled and the observations considered to all phases of the seismic cycle, on a fault interface that experiences earthquakes at multiple points in space and time. The motivating concept is that the recovery of rheological parameters could be greatly improved when considering that the constitutive laws for fault material must be able to reproduce all phases of the earthquake cycle, since it is the same physical material. Here, we (1) develop a timeseries analysis software that enables the efficient processing of large geodetic networks with long timeseries, allowing us to extract the relevant subduction-zone related signal in the observations, (2) formulate a forward model that, based on ancillary historical and seismic datasets as well as a candidate rheological model, simulates surface motion over multiple earthquake cycles, and (3) use a probabilistic inverse method to estimate the best-fitting rheological parameters given the postprocessed surface deformation timeseries and model uncertainties.
We validate the timeseries analysis software on the transient volcanic deformation of Long Valley Caldera, California, USA, before extracting the megathrust component of the surface observations on Northern Honshu Island, Japan. We then estimate the rheological properties of the Northern Japanese subduction zone using our inversion method, simultaneously producing time-varying estimates of kinematic coupling, slip deficit, and surface deformation. Our model predictions match the pre- and postseismic displacement timeseries of the 2011 Tohoku-oki earthquake well. On the steadily creeping part of the plate interface, we infer rate-dependent frictional parameters generally increasing with depth, but with second-order along-strike variation. Finally, we discuss the potential impact of our cycle-spanning, probabilistic inversion method on the field of subduction zone studies, and present possible avenues for further improvements to our framework.</p
Quantum Gravity and Laser Interferometry: Towards Observable Predictions
Get PDFUnderstanding quantum gravity remains one of the deepest challenges in modern physics, as direct experimental access to Planck-scale effects is beyond current technological reach. However, recent theoretical advances indicate that quantum fluctuations of spacetime may produce measurable effects in precision experiments, particularly near causal horizons. This opens new avenues for testing quantum gravity phenomena through high-precision measurement techniques. This dissertation develops multiple theoretical models to characterize these effects and examines their potential observational signatures in future gravitational wave interferometers.
We begin by investigating the role of quantum fluctuations in near-horizon geometries through the lens of the AdS/CFT correspondence, which provides a powerful framework for understanding the interplay between quantum field theory and general relativity via holographic principles. By modeling stochastic energy-momentum sources in Rindler-AdS spacetime, we demonstrate that vacuum fluctuations transform the Einstein equations into a Langevin-type stochastic differential equation, leading to potentially observable fluctuations in photon traversal times. Extending this approach to Minkowski spacetime, we establish a correspondence between gravitational shockwaves and fluid dynamics, showing that near-horizon perturbations satisfy an equation analogous to that governing incompressible fluids, thereby reinforcing the membrane paradigm and hydrodynamic analogies in the context of the fluid/gravity duality. Furthermore, we construct the covariant phase space of a spherically symmetric causal diamond in Minkowski spacetime, identifying two fundamental charges that govern its evolution. These results provide a foundation for quantizing causal horizons and understanding their microscopic degrees of freedom.
Building upon these theoretical developments, we further examine a related stochastic phenomenon: the gravitational wave memory background arising from the cumulative memory steps produced by supermassive black hole mergers. After reviewing the standard stochastic gravitational wave background, gravitational memory effects, and BMS symmetries, we model the stochastic memory background using a Brownian motion framework. We show that while the cumulative memory background initially appears above the sensitivity curve of space-based interferometers like LISA, the realistic subtraction of individually resolvable merger events substantially suppresses the residual signal, making its detection more challenging. This highlights the critical importance of source subtraction when evaluating the detectability of gravitational memory effects.
By bridging fundamental theory with experimental prospects, this dissertation contributes to the ongoing effort to uncover the quantum nature of spacetime through precision measurement techniques. Whether through detecting quantum spacetime fluctuations, gravitational memory backgrounds, or probing the symmetries of causal horizons, the pursuit of observable quantum gravity phenomena continues to expand the frontiers of both theory and experiment.</p
RNA-Mediated Toxicity In Neurodegeneration: The Mechanistic Role Of The C9ORF72 Repeat Expansion In ALS Molecular Pathogenesis
No full textThe G4C2 hexanucleotide repeat expansion in the first intron of the C9ORF72 gene is the most common genetic mutation linked to ALS, accounting for ~40 percent of familial and 10 percent of sporadic cases. Yet, its functional contribution to molecular pathogenesis remains unknown. The prevailing model is that this expansion leads to transcription of a novel RNA (C9-repeat RNA) that leads to disease either through its RNA product or translation of dipeptide repeat proteins it encodes (“gain-of-function”). However, recent attempts to degrade the C9-repeat RNA in several major clinical trials have failed to show any improvement in C9-ALS patients, raising questions about what role, if any, the C9-repeat RNA plays in ALS pathogenesis. Here, we demonstrate that the C9-repeat RNA is not detectable in C9-ALS patient-derived iPSNs or postmortem brain tissue. We show that transcription of the C9ORF72 gene initiates downstream of the G4C2 repeat sequence with the repeat expansion residing at a promoter-proximal region and displaying chromatin signatures of an enhancer. Because this region is GC-rich and has been reported to be preferentially methylated in C9-ALS patients, we explored whether this repeat expansion might lead to reduced C9ORF72 gene expression. We show that the C9-repeat is associated with reduced allele-specific expression of the C9ORF72 gene, consistent with the GC-rich features of the repeat expansion and previous reports of preferential DNA methylation in C9-ALS patients. Taken together, our findings challenge the prevailing gain-of-function models in C9-ALS and instead suggest that the repeat expansion region may function as a regulatory element that silences C9ORF72 expression from the mutant allele
Essays on Information Economics
Get PDFThis paper on information economics contains three chapters. In the first chapter, we study how to incentivize information acquisition in a principal-agent model. A principal hires an agent to collect information about a state. We study the optimal contract that incentivizes the agent to acquire the most precise information. In the second chapter, we study how to recover information in the selection model. We show that, given the selection rule and the observed selected outcome distribution, the entire outcome distribution can be characterized as the fixed point of an operator, which we prove to be a functional contraction. In the third chapter, we study how to implement randomized allocation rules with outcome-contingent transfers