137091 research outputs found
Sort by
Safe Control of Partially Unknown Systems Leveraging Efficient Reachability
No full textAutonomous systems operating in real-world conditions often have to contend with environmental disturbance behavior that is unknown a priori. We present a method for efficiently computing reachable sets for continuous-time systems with partially unknown dynamics. Our main assumption is that, given any hyperrectangle of states, lower and upper bounds for the unknown components are available. With this assumption, the theory of mixed monotone systems allows us to formulate an efficient method for computing hyperrectangular overapproximations of the reachable sets of the system.
We apply this formulation to a system navigating towards a goal region while avoiding unsafe regions. We derive a model predictive control scheme that avoids the unsafe region and ensures the system is always within reach of an a priori guaranteed safe region, thus ensuring feasibility until the goal is reachable. We explore this formulation further by considering multiple probability levels to increase performance.
We also consider the problem of tracking a reference trajectory for these systems. We modify the embedding system such that a single controlled trajectory corresponds to a controlled forward invariant interval tube around the reference, and utilize it in a runtime assurance mechanism that guarantees tracking of the reference trajectory within a desired threshold
Scalable and Provable Decision-Making for Large-Population Multi-Agent Systems in Complex Domains
No full textDecision-making algorithms with increased autonomy are critical for enabling future multi-agent systems. Unlike single-agent settings, multi-agent systems face severe scalability challenges due to inter-agent interactions. In particular, the joint policy and state spaces grow exponentially with the number of agents and the environment’s complexity, rendering classical approaches inadequate. This dissertation develops new formulations, frameworks, and algorithms to improve the efficiency and robustness of decision-making in large-scale multi-agent systems. The contributions address scalability along two key dimensions: the number of agents and the size of the state space.
For challenges associated with large-population teams, we employ a mean-field approximation, which reduces interactions among agents to those between a representative agent and the aggregate population distribution. We extend this framework to discrete state and action spaces by introducing entropy regularization to address the lack of regularity. To capture adversarial team interactions, we propose a zero-sum mean-field team game and provide a theoretical justification for the identical strategy simplification on the opponent team, enabling tractable analysis in competitive multi-team settings. Building on these foundations, we develop a scalable algorithm to learn competitive strategies for systems with hundreds of agents.
The study of large-population adversarial interactions led to the dynamic Defender Attacker Blotto game formulation, which generalizes static Blotto models to dynamic resource allocation over graphs. Through reachability analysis, we removed the classical assumption of instantaneous relocation and derived necessary and sufficient conditions for successful defense. These results bridge abstract allocation models with realistic robotic defense scenarios, where mobility constraints and network topology critically shape outcomes.
To address large state spaces, we develop a hierarchical decomposition framework for stochastic games. By leveraging the options framework, we construct a meta-game over coarse state representations, thereby reducing the complexity of computing Nash equilibria. We further enhance this approach with an automatic state aggregation method based on multi-timescale learning and tree-based abstractions, which adaptively partitions the state space to balance approximation accuracy with tractability, eliminating the need for hand-crafted abstractions.
Finally, we investigate decision-making under asymmetric information in hierarchical games. We introduce the notion of abstraction concealment, wherein agents strategically obscure their internal environment representations from adversaries. Formulated as a Bayesian game, the problem is solved using a bilinear program that computes perfect Bayesian equilibria. This represents one of the first principled treatments of deception through abstraction concealment, revealing how reduced representations shape equilibrium behaviors in adversarial settings.
Together, these contributions advance the scalability and robustness of multi-agent decision-making, bridging theory and algorithm design for hierarchical, large-population, and information-asymmetric settings
A Heterogeneous Integrated Photonic Platform for High Sensitivity Sensors
No full textHorizontal slot microdisk resonators provide a new pathway for integrated photonic sensing by combining ultra-high optical performance with strong interaction between light and analytes. In this work, a complete design and nanofabrication process was developed to realize these devices, which were validated through chemical and biomolecular sensing experiments. The approach demonstrated reliable detection of both refractive index changes and surface-adsorbed biomolecular layers, including the clinically relevant biomarker Troponin I. These results highlight the horizontal slot architecture as a powerful platform for lab-on-chip technologies with applications in healthcare diagnostics, environmental monitoring, and portable point-of-care systems
A Close Look at the Phase Behavior of Polymer Semiconductors Using Fast Calorimetry
No full textConjugated polymers have emerged as a promising class of semiconductor materials, thanks to their tunability and mechanical flexibility, among other desirable properties. Over the years, polymer semiconductors have made significant strides in applications like photovoltaics. However, these advancements often stem more from the vast array of organic moieties available to chemists than from a deep understanding of structure-property relationships. Instead of focusing on refining existing material systems, many innovations arise from novel chemical structures, which further complicates our grasp of these critical relationships.
The goal of this thesis is to establish new structure-property relationships for conjugated polymer semiconductors. Chapter 1 provides background information and motivation, introducing the implementation of fast scanning calorimetry and concluding with an outline of the thesis scope. Chapter 2 builds on this foundation by detailing the characterization methodologies and relevant material information. In Chapter 3, we apply fast scanning calorimetry to study both single-phase and two-phase commodity polymers, developing a protocol to track and quantify phase transition temperatures. This methodology is extended in Chapter 4 to semiflexible poly(3-hexylthiophene) (P3HT) derivatives, where heteroatom substitution is used to increase backbone rigidity and promote chain-extended structures. Chapter 5 continues this progression by further enhancing backbone rigidity and planarity, introducing the sanidic-like class of polymers, distinguished by their extended sidechains and ribbon-like morphology. The thesis concludes in Chapter 6 with a summary of key findings and an outlook on future research directions. Additional supporting data and discussion are presented in the appendix. Collectively, the insights from these studies aim to inform new design principles by highlighting the roles of backbone rigidity and sidechain ordering in the performance of polymer semiconductors
Improving Prediction for Disease-Associated Frameshift and Nonsense Mutations
No full textFrameshift and nonsense mutations account for approximately 8.4\% of disease-causing mutations but have received substantially less computational attention than missense variants. This research builds upon ENTPRISE-X and TransPPMP by incorporating protein language model embeddings and implementing class-aware loss functions to improve pathogenicity prediction, with particular focus on reducing false positive rates. We systematically evaluate three loss functions designed to address class imbalance: Focal Loss, Class-Balanced Loss, and Label-Distribution-Aware Margin (LDAM) Loss.
Our best-performing model, utilizing Class Balanced Loss combined with ESM-2 embeddings and mutation index features, achieves a Matthews Correlation Coefficient (MCC) of 0.697, F-score of 0.717, sensitivity of 0.817, and specificity of 0.963 on the VEST-indel test set. This represents a 4.2\% improvement in MCC over the current state-of-the-art TransPPMP method (MCC: 0.669, Specificity: 0.939), while achieving a 59\% reduction in false positive rate (from 6.1\% to 3.7\%) and maintaining strong sensitivity.
An alternative configuration using Class Balanced Focal Loss achieves even higher specificity (0.976) for applications where minimizing false positives is paramount, at the cost of reduced sensitivity (0.732). These results demonstrate that strategic loss function selection can meaningfully reduce false positives in mutation pathogenicity prediction without compromising overall performance, offering practical value for clinical variant interpretation where each false positive triggers expensive confirmatory testing and patient burden
Developing a Material Property Library for Fused Deposition Modeling 3D printed parts to Support the Design of Assistive Technology
No full textAdditive manufacturing, enables the rapid and affordable production of customized, low-volume assistive technology (AT) devices. However, a major challenge that persists is that the mechanical behavior of 3D printed parts is difficult to predict due to their anisotropic nature and the strong influence of printing parameters, such as infill pattern and density. These inconsistencies make it difficult to use finite element analysis (FEA) tools effectively. In most CAD environments, simulations rely on idealized or bulk material properties that don’t represent the actual printed geometry or internal structure.
This research aims to address this gap by developing a material property library that accounts for the anisotropic nature of FDM components, with a specific focus on supporting the design of low-volume assistive technology products. The primary objective is to empirically characterize common 3D printing materials (PLA, and PETG) by testing specimens fabricated with a range of infill patterns and infill densities to compile a structured material property library for direct use in CAD-based simulations.
By providing AT developers with this data, this work aims to establish a data-driven framework for designing FDM-based assistive technologies (AT), reduce the need to rely on over-engineered designs, and ultimately support the creation of safer and efficient assistive technology products
In-Situ Monitoring and Solubility Modeling of Sodium Phosphate Using Online PAT Tools
No full textThe Hanford site contains 56 million gallons of radioactive waste that was produced starting in the Manhattan project and ending in the cold war. The Hanford tank waste consists of supernatant liquid, salt cake, and sludge. The alkaline liquid phase is rich
in salts, including sodium phosphate. The blockages have previously caused delays in operations and an approximate cost of 40 million dollars from the replacement of clogged
underground piping. The crystallization of material that has previously clogged pipelines at Hanford may potentially crystallize during processing at the Hanford Waste Treatment Plant. However, optical spectroscopy techniques, such as infrared and Raman spectroscopy, have demonstrated potential for online monitoring of solutions and slurries and may be able to detect inadvertent crystallization quickly or before crystallization has occurred. In this work, in-situ and ex-situ measurements are used to identify the hydrate of sodium phosphate that crystallizes at high pH and to build a solubility model based on the Pitzer framework.
In this study, Process Analytical Technology (PAT) tools were employed to monitor different forms of sodium phosphate that crystallize from cooling crystallization at different concentrations, temperatures, and cooling rates. A Mettler Toledo EasyViewer probe visually inspected the crystals, facilitating differentiation of morphologies. Raman spectroscopy
was used to analyze the solid and solution phases; Raman spectroscopy can distinguish phosphate and hydrogen phosphate anions in the solution phase while also detecting
solid crystals. Attenuated Total Reflectance – Fourier Transform Infrared (ATR-FTIR) spectroscopy was expected to measure the solution phase only; ATR-FTIR absorbance
varies directly with solute concentration, offering insights into dissolved phosphate ions.
The sodium phosphate crystalline phases formed were characterized using Differential Scanning Calorimetry (DSC), Thermogravimetric Analysis (TGA), Raman Spectroscopy, and X-ray Diffraction (XRD). DSC and TGA results indicated the formation of hydrated trisodium phosphate species, and Raman spectroscopy data, both in situ and ex situ, were consistent with these thermal analyses. However, XRD analysis suggests that the crystallized
phases may include trisodium phosphate hemihydrate and disodium hydrogen phosphate. Raman spectroscopy, collected in air, showed PO3−4 peaks at ∼942 and 1006 cm−1
and no detectable HPO2−4 features. Because the crystals progressively lose water under ambient conditions, they trend toward lower hydrate forms over time. Therefore, in situ Raman and ATR-FTIR measurements are essential for capturing the true crystalline and solution species at the moment of formation before washing and air-drying alters them.
The sodium phosphate crystalline phases formed were characterized using Differential Scanning Calorimetry (DSC), Thermogravimetric Analysis (TGA), Raman Spectroscopy, and X-ray Diffraction (XRD). DSC and TGA results indicated the formation of hydrated trisodium phosphate species, and Raman spectroscopy data, both in situ and ex situ, were consistent with these thermal analyses. However, XRD analysis suggests that the crystallized
phases may include trisodium phosphate hemihydrate and disodium hydrogen phosphate. Raman spectroscopy, collected in air, showed PO3−4 peaks at ∼942 and 1006 cm−1
and no detectable HPO2−4 features. Because the crystals progressively lose water under ambient conditions, they trend toward lower hydrate forms over time. Therefore, in situ Raman and ATR-FTIR measurements are essential for capturing the true crystalline and solution species at the moment of formation before washing and air-drying alters them.
To interpret and generalize the experimental solubility behavior of sodium phosphate in alkaline media, a thermodynamic model was developed to relate solubility to temperature and ionic interactions. Experimental solubility data across a range of temperatures (25–55◦C) and fixed sodium hydroxide concentration (3 molal) were used to parameterize a thermodynamic solubility model based on the Pitzer framework. The resulting model captured the complex interactions within the Na3PO4-NaOH-H2O system, demonstrating good accuracy within the studied temperature range.
This research demonstrates that integrating in-situ ATR-FTIR and Raman spectroscopy with EasyViewer provides a comprehensive, real-time picture of phosphate crystallization
under strongly alkaline conditions. The EasyViewer offered continuous visualization of crystal size and shape evolution. ATR-FTIR tracked the solution phase PO3−4 band (∼1003
cm−1) with baseline-corrected absorbance that is linearly proportional to concentration, allowing us to quantify dissolved phosphate throughout each heating-cooling cycle. Simultaneously, Raman spectroscopy captured both solid and liquid signatures: the PO3−4 v1 stretch shifted from 936 cm−1 in clear solution to 942 cm−1 once crystals appeared, and a 1006 cm−1 v3 band appeared only in the presence of a slurry, evidence that a trisodium phosphate hydrate was the phase crystallized. The absence of HPO2−4 Raman bands further confirmed that the solid crystallized consisted only PO3−4 species.M.S.Chemical and Biomolecular Engineerin
Deep Learning Characterization and Mechanical Ranking of Microstructure Features in Geomaterials
No full textUnderstanding the analytical relationship between fabric descriptors (i.e., microstructure descriptors) and stiffness tensors has been a long-standing challenge in geomechanics. This doctoral research aims to address this challenge by exploring three key research questions.
The first research question focuses on calculating 3D fabric descriptors from 2D images by using deep learning (DL) models. The performance of a pruned ACS-VGG19 network is assessed under different training metrics and configurations of trainable and fixed convolutional layers. The optimal model configuration utilizes the MSE loss function and fully trainable convolutional layers, achieving a Mean Absolute Percentage Error (MAPE) of 2 to 5\% for aggregate size, aspect ratios, and solidity. Computational costs increase linearly with the number of images extracted per direction, but performance improvements are marginal beyond a single image per direction.
The second research question is to understand the relative importance of microstructure features on local field variables, such as the stress field. Here, 2D composite materials made of a solid matrix and cracks are modeled numerically with the Finite Element Method (FEM) and cohesive zone models (CZM). A Non-Linear Variational Autoencoder (NLVAE) is proposed with a skew-normal distribution sampling and correlation penalties to improve latent feature disentanglement. The NLVAE effectively captures stress concentrations and reconstructs stress fields with consistent Normalized Mean Square Error (NMSE) across different stress components. Dynamic Time Warping (DTW) analysis reveals correlations between stress latent features and crack network descriptors, such as connectivity, path length, eccentricity, and centrality. Significant stress latent features and microstructure descriptors vary with boundary conditions, but certain latent variables consistently emerge as significant across different descriptors and loading paths. Notably, the distributional shape, tail behavior, and symmetry of microstructure descriptor distributions have more influence on the stress field than basic measures of central tendency and spread.
The final research question aims to explain why classic homogenization schemes break down in cracked solids. The information flow from microstructure variations to stiffness changes is measured. A support vector machine (SVM) with feature selection using mutual information (MI) and Analysis of Variance (ANOVA) highlights the influence of individual fabric descriptors on the breakdown of the Mori-Tanaka model.
This thesis explores the relative importance of statistical geometric descriptors in biphase composites for the reconstruction of 3D microstructure images, the estimation of the stress field, and the calculation of effective stiffness. Deep learning algorithms have been developed to discover microstructure self-organization patterns and to explain micromechanical phenomena. The results not only fill existing gaps in microstructure characterization, influential fabric recognition, and homogenization theory, but also offer numerous opportunities for future research, such as the effect of data preparation, extensions to 3D composites, metrics between sequences of statistical datasets, and extrapolation from DL.Ph.D.Civil Engineerin
Cryogenic Lunar Sample Return
No full textPresented at AIAA SciTech 2026The discovery of lunar water-ice volatiles in shadowed regions of the Moon potentially offers abundant scientific information about lunar history and resources for future lunar missions. These volatiles must be preserved at cryogenic temperatures to retain their scientific integrity. Current lunar sample return plans include a stop at the Lunar Gateway space station and transport the samples within crew vehicles. This method prolongs the samples’ transit time and imposes crew-safe handling restrictions which complicates the cryogenic storage requirement. A direct, uncrewed lunar sample return would bypass both constraints and ultimately allow for a potentially simpler mission architecture. This study conceptually analyzes a mission to directly return surface volatiles from the Lunar South Pole focusing on the launch and return trajectory, Earth reentry, and payload design. It utilizes a derivative of the Mars Ascent Vehicle for launch from the Moon and a mechanically deployable heat shield for reentry. A passive cryogenics system utilizing liquid nitrogen or neon is used for cooling. It was found that such a mission could return to Earth up to 6 kg of samples frozen at 77 K